Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1...

109
Transport Layer 3-1 Chapter 3 Transport Layer Computer Networking: A Top Down Approach 5 th edition. Jim Kurose, Keith Ross Addison-Wesley, April 2009. All material copyright 1996-2009 J.F Kurose and K.W. Ross, All Rights Reserved

Transcript of Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1...

Page 1: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-1

Chapter 3Transport Layer

Computer NetworkingA Top Down Approach5th editionJim Kurose Keith RossAddison-Wesley April2009

All material copyright 1996-2009JF Kurose and KW Ross All Rights Reserved

Transport Layer 3-2

Chapter 3 Transport LayerOur goals understand principles

behind transportlayer services multiplexingdemultipl

exing reliable data transfer flow control congestion control

learn about transportlayer protocols in theInternet UDP connectionless

transport TCP connection-oriented

transport TCP congestion control

Transport Layer 3-3

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-4

Transport services and protocols provide logical communication

between app processesrunning on different hosts

transport protocols run in endsystems send side breaks app

messages into segmentspasses to network layer

rcv side reassemblessegments into messagespasses to app layer

more than one transportprotocol available to apps Internet TCP and UDP

applicationtransportnetworkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-5

Transport vs network layer

network layer logicalcommunicationbetween hosts

transport layer logicalcommunicationbetween processes relies on enhances

network layer services

Household analogy12 kids sending letters

to 12 kids processes = kids app messages = letters

in envelopes hosts = houses transport protocol =

Ann and Bill network-layer protocol

= postal service

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 2: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-2

Chapter 3 Transport LayerOur goals understand principles

behind transportlayer services multiplexingdemultipl

exing reliable data transfer flow control congestion control

learn about transportlayer protocols in theInternet UDP connectionless

transport TCP connection-oriented

transport TCP congestion control

Transport Layer 3-3

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-4

Transport services and protocols provide logical communication

between app processesrunning on different hosts

transport protocols run in endsystems send side breaks app

messages into segmentspasses to network layer

rcv side reassemblessegments into messagespasses to app layer

more than one transportprotocol available to apps Internet TCP and UDP

applicationtransportnetworkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-5

Transport vs network layer

network layer logicalcommunicationbetween hosts

transport layer logicalcommunicationbetween processes relies on enhances

network layer services

Household analogy12 kids sending letters

to 12 kids processes = kids app messages = letters

in envelopes hosts = houses transport protocol =

Ann and Bill network-layer protocol

= postal service

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 3: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-3

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-4

Transport services and protocols provide logical communication

between app processesrunning on different hosts

transport protocols run in endsystems send side breaks app

messages into segmentspasses to network layer

rcv side reassemblessegments into messagespasses to app layer

more than one transportprotocol available to apps Internet TCP and UDP

applicationtransportnetworkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-5

Transport vs network layer

network layer logicalcommunicationbetween hosts

transport layer logicalcommunicationbetween processes relies on enhances

network layer services

Household analogy12 kids sending letters

to 12 kids processes = kids app messages = letters

in envelopes hosts = houses transport protocol =

Ann and Bill network-layer protocol

= postal service

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 4: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-4

Transport services and protocols provide logical communication

between app processesrunning on different hosts

transport protocols run in endsystems send side breaks app

messages into segmentspasses to network layer

rcv side reassemblessegments into messagespasses to app layer

more than one transportprotocol available to apps Internet TCP and UDP

applicationtransportnetworkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-5

Transport vs network layer

network layer logicalcommunicationbetween hosts

transport layer logicalcommunicationbetween processes relies on enhances

network layer services

Household analogy12 kids sending letters

to 12 kids processes = kids app messages = letters

in envelopes hosts = houses transport protocol =

Ann and Bill network-layer protocol

= postal service

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 5: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-5

Transport vs network layer

network layer logicalcommunicationbetween hosts

transport layer logicalcommunicationbetween processes relies on enhances

network layer services

Household analogy12 kids sending letters

to 12 kids processes = kids app messages = letters

in envelopes hosts = houses transport protocol =

Ann and Bill network-layer protocol

= postal service

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 6: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-6

Internet transport-layer protocols

reliable in-orderdelivery (TCP) congestion control flow control connection setup

unreliable unordereddelivery UDP no-frills extension of

ldquobest-effortrdquo IP services not available

delay guarantees bandwidth guarantees

applicationtransportnetworkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

networkdata linkphysical

applicationtransportnetworkdata linkphysical

logical end-end transport

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 7: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-7

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 8: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-8

Multiplexingdemultiplexing

application

transport

network

link

physical

P1 application

transport

network

link

physical

application

transport

network

link

physical

P2P3 P4P1

host 1 host 2 host 3

= process= socket

delivering received segmentsto correct socket

Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)

Multiplexing at send host

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 9: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-9

How demultiplexing works host receives IP datagrams

each datagram has sourceIP address destination IPaddress

each datagram carries 1transport-layer segment

each segment has sourcedestination port number

host uses IP addresses amp portnumbers to direct segment toappropriate socket

source port dest port

32 bits

applicationdata

(message)

other header fields

TCPUDP segment format

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 10: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-10

Connectionless demultiplexing

Create sockets with portnumbers

DatagramSocket mySocket1 = newDatagramSocket(12534)

DatagramSocket mySocket2 = newDatagramSocket(12535)

UDP socket identified bytwo-tuple

(dest IP address dest port number)

When host receives UDPsegment checks destination port

number in segment directs UDP segment to

socket with that portnumber

IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 11: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-11

Connectionless demux (cont)

DatagramSocket serverSocket = new DatagramSocket(6428)

ClientIPB

P2

client IP A

P1P1P3

serverIP C

SP 6428DP 9157

SP 9157DP 6428

SP 6428DP 5775

SP 5775DP 6428

SP provides ldquoreturn addressrdquo

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 12: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-12

Connection-oriented demux

TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number

receiving host uses allfour values to directsegment to appropriatesocket

Server host may supportmany simultaneous TCPsockets each socket identified by

its own 4-tuple Web servers have

different sockets foreach connecting client non-persistent HTTP will

have different socket foreach request

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 13: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-13

Connection-oriented demux(cont)

ClientIPB

P1

client IP A

P1P2P4

serverIP C

SP 9157DP 80

SP 9157DP 80

P5 P6 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 14: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-14

Connection-oriented demuxThreaded Web Server

ClientIPB

P1

client IP A

P1P2

serverIP C

SP 9157DP 80

SP 9157DP 80

P4 P3

D-IPCS-IP AD-IPC

S-IP B

SP 5775DP 80

D-IPCS-IP B

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 15: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-15

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 16: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-16

UDP User Datagram Protocol [RFC 768]

ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol

ldquobest effortrdquo service UDPsegments may be lost delivered out of order

to app connectionless

no handshaking betweenUDP sender receiver

each UDP segmenthandled independentlyof others

Why is there a UDP no connection

establishment (which canadd delay)

simple no connection stateat sender receiver

small segment header no congestion control UDP

can blast away as fast asdesired

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 17: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-17

UDP more often used for streaming

multimedia apps loss tolerant rate sensitive

other UDP uses DNS SNMP

reliable transfer over UDPadd reliability atapplication layer application-specific

error recovery

source port dest port

32 bits

Applicationdata

(message)

UDP segment format

length checksumLength in

bytes of UDPsegmentincluding

header

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 18: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-18

UDP checksum

Sender treat segment contents as

sequence of 16-bitintegers

checksum addition (1rsquoscomplement sum) ofsegment contents

sender puts checksumvalue into UDP checksumfield

Receiver compute checksum of

received segment check if computed checksum

equals checksum field value NO - error detected YES - no error detected

But maybe errorsnonetheless More laterhellip

Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 19: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-19

Internet Checksum Example Note

When adding numbers a carryout from themost significant bit needs to be added to theresult

Example add two 16-bit integers

1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1

1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1

wraparound

sumchecksum

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 20: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-20

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 21: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-21

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 22: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-22

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 23: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-23

Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics

characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 24: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-24

Reliable data transfer getting started

sendside

receiveside

rdt_send() called from above(eg by app) Passed data to

deliver to receiver upper layer

udt_send() called by rdtto transfer packet over

unreliable channel to receiver

rdt_rcv() called when packetarrives on rcv-side of channel

deliver_data() called byrdt to deliver data to upper

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 25: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-25

Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of

reliable data transfer protocol (rdt) consider only unidirectional data transfer

but control info will flow on both directions use finite state machines (FSM) to specify

sender receiver

state1

state2

event causing state transitionactions taken on state transition

state when in thisldquostaterdquo next state

uniquely determinedby next event

eventactions

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 26: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-26

Rdt10 reliable transfer over a reliable channel

underlying channel perfectly reliable no bit errors no loss of packets

separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel

Wait forcall fromabove packet = make_pkt(data)

udt_send(packet)

rdt_send(data)extract (packetdata)deliver_data(data)

Wait forcall from

below

rdt_rcv(packet)

sender receiver

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 27: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-27

Rdt20 channel with bit errors

underlying channel may flip bits in packet checksum to detect bit errors

the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender

that pkt received OK negative acknowledgements (NAKs) receiver explicitly

tells sender that pkt had errors sender retransmits pkt on receipt of NAK

new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 28: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-28

rdt20 FSM specification

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

belowsender

receiverrdt_send(data)

Λ

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 29: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-29

rdt20 operation with no errors

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 30: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-30

rdt20 error scenario

Wait forcall fromabove

snkpkt = make_pkt(data checksum)udt_send(sndpkt)

extract(rcvpktdata)deliver_data(data)udt_send(ACK)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)

udt_send(NAK)

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Wait forACK or

NAK

Wait forcall from

below

rdt_send(data)

Λ

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 31: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-31

rdt20 has a fatal flaw

What happens ifACKNAK corrupted

sender doesnrsquot know whathappened at receiver

canrsquot just retransmitpossible duplicate

Handling duplicates sender retransmits current

pkt if ACKNAK garbled sender adds sequence

number to each pkt receiver discards (doesnrsquot

deliver up) duplicate pkt

Sender sends one packet then waits for receiver response

stop and wait

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 32: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-32

rdt21 sender handles garbled ACKNAKs

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

Wait forACK orNAK 0 udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)

Wait for call 1 from

above

Wait forACK orNAK 1

ΛΛ

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 33: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-33

rdt21 receiver handles garbled ACKNAKs

Wait for0 frombelow

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

Wait for1 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)

sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)

sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 34: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-34

rdt21 discussion

Sender seq added to pkt two seq rsquos (01) will

suffice Why must check if received

ACKNAK corrupted twice as many states

state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq

Receiver must check if received

packet is duplicate state indicates whether

0 or 1 is expected pktseq

note receiver can notknow if its lastACKNAK received OKat sender

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 35: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-35

rdt22 a NAK-free protocol

same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt

received OK receiver must explicitly include seq of pkt being ACKed

duplicate ACK at sender results in same action asNAK retransmit current pkt

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 36: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-36

rdt22 sender receiver fragments

Wait forcall 0 from

above

sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)

rdt_send(data)

udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

Wait forACK

0sender FSM

fragment

Wait for0 frombelow

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)

rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))

udt_send(sndpkt)receiver FSM

fragment

Λ

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 37: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-37

rdt30 channels with errors and loss

New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs

retransmissions will beof help but not enough

Approach sender waitsldquoreasonablerdquo amount oftime for ACK

retransmits if no ACKreceived in this time

if pkt (or ACK) just delayed(not lost) retransmission will be

duplicate but use of seqrsquos already handles this

receiver must specify seq of pkt being ACKed

requires countdown timer

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 38: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-38

rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

Waitfor

ACK0

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )

Wait forcall 1 from

above

sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer

rdt_send(data)

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)

rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )

rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)

stop_timerstop_timer

udt_send(sndpkt)start_timer

timeout

udt_send(sndpkt)start_timer

timeout

rdt_rcv(rcvpkt)

Wait forcall 0from

above

Waitfor

ACK1

Λrdt_rcv(rcvpkt)

ΛΛ

Λ

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 39: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-39

rdt30 in action

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 40: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-40

rdt30 in action

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 41: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-41

Performance of rdt30

rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet

U sender utilization ndash fraction of time sender busy sending

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources

dsmicrosecon8bps10

bits80009

===R

Ldtrans

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 42: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-42

rdt30 stop-and-wait operation

first packet bit transmitted t = 0

sender receiver

RTT

last packet bit transmitted t = L R

first packet bit arriveslast packet bit arrives sendACK

ACK arrives send nextpacket t = RTT + L R

U sender

= 008

30008 = 000027

microsec

onds

L R

RTT + L R =

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 43: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-43

Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-

be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver

Two generic forms of pipelined protocols go-Back-Nselective repeat

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 44: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-44

Pipelining increased utilization

first packet bit transmitted t = 0

sender receiver

RTT

last bit transmitted t = L R

first packet bit arriveslast packet bit arrives send ACK

ACK arrives send nextpacket t = RTT + L R

last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK

U sender

= 024

30008 = 00008

microsecon

ds

3 L R

RTT + L R =

Increase utilizationby a factor of 3

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 45: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-45

Pipelining Protocols

Go-back-N overview sender up to N

unACKed pkts inpipeline

receiver only sendscumulative ACKs doesnrsquot ACK pkt if

therersquos a gap sender has timer for

oldest unACKed pkt if timer expires

retransmit all unACKedpackets

Selective Repeat overview sender up to N unACKed

packets in pipeline receiver ACKs individual

pkts sender maintains timer

for each unACKed pkt if timer expires retransmit

only unACKed packet

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 46: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-46

Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed

ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)

timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 47: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-47

GBN sender extended FSM

Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])

timeout

rdt_send(data)

if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)

base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer

rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)

base=1nextseqnum=1

rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)

Λ

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 48: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-48

GBN receiver extended FSM

ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum

out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq

Wait

udt_send(sndpkt)default

rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)

extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++

expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)

Λ

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 49: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-49

GBN inaction

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 50: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-50

Selective Repeat

receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery

to upper layer sender only resends pkts for which ACK not

received sender timer for each unACKed pkt

sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 51: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-51

Selective repeat sender receiver windows

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 52: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-52

Selective repeat

data from above if next available seq in

window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]

mark pkt n as received if n smallest unACKed pkt

advance window base tonext unACKed seq

senderpkt n in [rcvbase rcvbase+N-1]

send ACK(n) out-of-order buffer in-order deliver (also

deliver buffered in-orderpkts) advance window tonext not-yet-received pkt

pkt n in [rcvbase-Nrcvbase-1]

ACK(n)otherwise ignore

receiver

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 53: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-53

Selective repeat in action

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 54: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-54

Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3

receiver sees nodifference in twoscenarios

incorrectly passesduplicate data as newin (a)

Q what relationshipbetween seq sizeand window size

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 55: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-55

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 56: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-56

TCP Overview RFCs 793 1122 1323 2018 2581

full duplex data bi-directional data flow

in same connection MSS maximum segment

size connection-oriented

handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange

flow controlled sender will not

overwhelm receiver

point-to-point one sender one receiver

reliable in-order bytesteam no ldquomessage boundariesrdquo

pipelined TCP congestion and flow

control set window size send amp receive buffers

socket

door

TCP

send buffer

TCP

receive buffer

socket

door

segment

application

writes data

application

reads data

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 57: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-57

TCP segment structure

source port dest port

32 bits

applicationdata

(variable length)

sequence numberacknowledgement number

Receive windowUrg data pointerchecksum

FSRPAUheadlen

notused

Options (variable length)

URG urgent data (generally not used)

ACK ACK valid

PSH push data now(generally not used)

RST SYN FINconnection estab(setup teardown

commands)

bytes rcvr willingto accept

countingby bytes of data(not segments)

Internetchecksum

(as in UDP)

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 58: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-58

TCP seq rsquos and ACKsSeq rsquos

byte streamldquonumberrdquo of firstbyte in segmentrsquosdata

ACKs seq of next byte

expected fromother side

cumulative ACKQ how receiver handles

out-of-order segments A TCP spec doesnrsquot

say - up toimplementer

Host A Host B

Seq=42 ACK=79 data = lsquoCrsquo

Seq=79 ACK=43 data = lsquoCrsquo

Seq=43 ACK=80

Usertypes

lsquoCrsquo

host ACKsreceipt

of echoedlsquoCrsquo

host ACKsreceipt oflsquoCrsquo echoes

back lsquoCrsquo

timesimple telnet scenario

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 59: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-59

TCP Round Trip Time and Timeout

Q how to set TCPtimeout value

longer than RTT but RTT varies

too short prematuretimeout unnecessary

retransmissions too long slow reaction

to segment loss

Q how to estimate RTT SampleRTT measured time from

segment transmission until ACKreceipt ignore retransmissions

SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent

measurements not justcurrent SampleRTT

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 60: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-60

TCP Round Trip Time and Timeout

EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT

Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 61: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-61

Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr

100

150

200

250

300

350

1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106

time (seconnds)

RT

T

(m

illise

co

nd

s)

SampleRTT Estimated RTT

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 62: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-62

TCP Round Trip Time and Timeout

Setting the timeout EstimtedRTT plus ldquosafety marginrdquo

large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from

EstimatedRTT

TimeoutInterval = EstimatedRTT + 4DevRTT

DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|

(typically β = 025)

Then set timeout interval

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 63: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-63

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 64: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-64

TCP reliable data transfer

TCP creates rdtservice on top of IPrsquosunreliable service

pipelined segments cumulative ACKs TCP uses single

retransmission timer

retransmissions aretriggered by timeout events duplicate ACKs

initially considersimplified TCP sender ignore duplicate ACKs ignore flow control

congestion control

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 65: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-65

TCP sender eventsdata rcvd from app create segment with

seq seq is byte-stream

number of first databyte in segment

start timer if notalready running (thinkof timer as for oldestunACKed segment)

expiration intervalTimeOutInterval

timeout retransmit segment

that caused timeout restart timer ACK rcvd if acknowledges

previously unACKedsegments update what is known to

be ACKed start timer if there are

outstanding segments

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 66: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-66

TCPsender(simplified)

NextSeqNum = InitialSeqNum SendBase = InitialSeqNum

loop (forever) switch(event)

event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)

event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer

end of loop forever

Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 67: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-67

TCP retransmission scenariosHost A

Seq=100 20 bytes data

ACK=100

timepremature timeout

Host B

Seq=92 8 bytes data

ACK=120

Seq=92 8 bytes data

Seq=

92 t

imeo

ut

ACK=120

Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

lost ACK scenario

Host B

X

Seq=92 8 bytes data

ACK=100

time

Seq=

92 t

imeo

utSendBase

= 100

SendBase= 120

SendBase= 120

Sendbase= 100

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 68: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-68

TCP retransmission scenarios (more)Host A

Seq=92 8 bytes data

ACK=100

loss

tim

eout

Cumulative ACK scenario

Host B

X

Seq=100 20 bytes data

ACK=120

time

SendBase= 120

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 69: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-69

TCP ACK generation [RFC 1122 RFC 2581]

Event at Receiver

Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed

Arrival of in-order segment withexpected seq One other segment has ACK pending

Arrival of out-of-order segmenthigher-than-expect seq Gap detected

Arrival of segment that partially or completely fills gap

TCP Receiver action

Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK

Immediately send single cumulative ACK ACKing both in-order segments

Immediately send duplicate ACK indicating seq of next expected byte

Immediate send ACK provided thatsegment starts at lower end of gap

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 70: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-70

Fast Retransmit

time-out period oftenrelatively long long delay before

resending lost packet detect lost segments

via duplicate ACKs sender often sends

many segments back-to-back

if segment is lost therewill likely be manyduplicate ACKs for thatsegment

If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend

segment before timerexpires

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 71: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-71

Host A

tim

eout

Host B

time

X

resend seq X2

seq x1seq x2seq x3seq x4seq x5

ACK x1

ACK x1ACK x1ACK x1

tripleduplicate

ACKs

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 72: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-72

event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y

Fast retransmit algorithm

a duplicate ACK for already ACKed segment

fast retransmit

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 73: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-73

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 74: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-74

TCP Flow Control

receive side of TCPconnection has areceive buffer

speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate

app process may beslow at reading frombuffer

sender wonrsquot overflowreceiverrsquos buffer by

transmitting toomuch

too fast

flow control

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 75: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-75

TCP Flow control how it works

(suppose TCP receiverdiscards out-of-ordersegments)

unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -

LastByteRead]

receiver advertisesunused buffer space byincluding rwnd value insegment header

sender limits ofunACKed bytes to rwnd guarantees receiverrsquos

buffer doesnrsquot overflow

IPdatagrams

TCP data(in buffer)

(currently)unused buffer

space

applicationprocess

rwndRcvBuffer

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 76: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-76

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 77: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-77

TCP Connection ManagementRecall TCP sender receiver

establish ldquoconnectionrdquobefore exchanging datasegments

initialize TCP variables seq s buffers flow control

info (eg RcvWindow) client connection initiator Socket clientSocket = new

Socket(hostnameportnumber)

server contacted by client Socket connectionSocket =

welcomeSocketaccept()

Three way handshakeStep 1 client host sends TCP

SYN segment to server specifies initial seq no data

Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq

Step 3 client receives SYNACK

replies with ACK segmentwhich may contain data

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 78: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-78

TCP Connection Management (cont)

Closing a connectionclient closes socket

clientSocketclose()

Step 1 client end systemsends TCP FIN controlsegment to server

Step 2 server receivesFIN replies with ACKCloses connection sendsFIN

client

FIN

server

ACK

ACK

FIN

close

close

closed

tim

ed w

ait

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 79: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-79

TCP Connection Management (cont)

Step 3 client receives FINreplies with ACK

Enters ldquotimed waitrdquo -will respond with ACKto received FINs

Step 4 server receivesACK Connection closed

Note with smallmodification can handlesimultaneous FINs

client

FIN

server

ACK

ACK

FIN

closing

closing

closed

tim

ed w

ait

closed

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 80: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-80

TCP Connection Management (cont)

TCP clientlifecycle

TCP serverlifecycle

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 81: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-81

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 82: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-82

Principles of Congestion Control

Congestion informally ldquotoo many sources sending too much

data too fast for network to handlerdquo different from flow control manifestations

lost packets (buffer overflow at routers) long delays (queueing in router buffers)

a top-10 problem

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 83: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-83

Causescosts of congestion scenario 1

two senders tworeceivers

one routerinfinite buffers

no retransmission

large delayswhen congested

maximumachievablethroughput

unlimited sharedoutput link buffers

Host Aλin original data

Host B

λout

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 84: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-84

Causescosts of congestion scenario 2

one router finite buffers sender retransmission of lost packet

finite shared outputlink buffers

Host A λin originaldata

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 85: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-85

Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss

retransmission of delayed (not lost) packet makes larger(than perfect case) for same

λin

λout=

λin

λoutgtλ

inλ

out

ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt

R2

R2λin

λ out

b

R2

R2λin

λ out

a

R2

R2λin

λ out

c

R4

R3

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 86: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-86

Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit

λin

Q what happens asand increase λ

in

finite shared outputlink buffers

Host Aλin original data

Host B

λout

λin original data plusretransmitted data

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 87: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-87

Causescosts of congestion scenario 3

another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission

capacity used for that packet was wasted

HostA

HostB

λout

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 88: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-88

Approaches towards congestion control

end-end congestioncontrol

no explicit feedback fromnetwork

congestion inferred fromend-system observed lossdelay

approach taken by TCP

network-assistedcongestion control

routers provide feedbackto end systems single bit indicating

congestion (SNADECbit TCPIP ECNATM)

explicit rate sendershould send at

two broad approaches towards congestion control

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 89: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-89

Case study ATM ABR congestion control

ABR available bit rate ldquoelastic servicerdquo if senderrsquos path

ldquounderloadedrdquo sender should use

available bandwidth if senderrsquos path

congested sender throttled to

minimum guaranteedrate

RM (resource management)cells

sent by sender interspersedwith data cells

bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate

(mild congestion) CI bit congestion

indication RM cells returned to sender by

receiver with bits intact

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 90: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-90

Case study ATM ABR congestion control

two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path

EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI

bit in returned RM cell

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 91: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-91

Chapter 3 outline

31 Transport-layerservices

32 Multiplexing anddemultiplexing

33 Connectionlesstransport UDP

34 Principles ofreliable data transfer

35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management

36 Principles ofcongestion control

37 TCP congestioncontrol

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 92: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-92

TCP congestion control Goal TCP sender should transmit as fast as possible

but without congesting network Q how to find rate just below congestion level

Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not

congested so increase sending rate lost segment assume loss due to congested

network so decrease sending rate

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 93: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-93

TCP congestion control bandwidth probing

ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available

bandwidth is changing depending on other connections innetwork)

ACKs being received so increase rate

X

X

XX

X loss so decrease rate

send

ing

rate

time

Q how fast to increasedecrease

TCPrsquosldquosawtoothrdquobehavior

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 94: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-94

TCP Congestion Control details

Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie

cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)

Sending rate or throughput roughly

Sender controls cwnd dynamically as afunction of perceived network congestion

rate = cwnd

RTT bytessec

LastByteSent-LastByteAcked le cwnd

cwndbytes

RTT

ACK(s)

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 95: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-95

TCP Congestion Control more details

segment loss eventreducing cwnd

timeout no responsefrom receiver cut cwnd to 1

3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less

aggressively than ontimeout

ACK received increasecwnd

Two phases slowstart

increase exponentiallyfast (despite name) atconnection start orfollowing timeout

congestion avoidance increase linearly

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 96: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-96

TCP Slow Start when connection begins cwnd =

1 MSS example MSS = 500 bytes

amp RTT = 200 msec initial rate = 20 kbps

available bandwidth may be gtgtMSSRTT desirable to quickly ramp up

to respectable rate increase rate exponentially

until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd

by 1 for every ACK received

Host A

one segment

RTT

Host B

time

two segments

four segments

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 97: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-97

Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when

congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion

avoidance phase

slow start timeout

ssthresh = cwnd2cwnd = 1 MSS

dupACKcount = 0retransmit missing segment timeout

ssthresh = cwnd2 cwnd = 1 MSS

dupACKcount = 0retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKdupACKcount++duplicate ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion

avoidance

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 98: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-98

TCP congestion avoidance

Goal Approach possiblecongestion slower than inslowstart

When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received

ACKs increase cwndby 1 MSS per RTTadditive increase

loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease

AIMD

AIMD Additive IncreaseMultiplicative Decrease

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 99: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-99

TCP congestion control FSM overview

slow start

congestionavoidance

fastrecovery

cwnd gt ssthresh

losstimeout

losstimeout

new ACK loss3dupACK

loss3dupACK

losstimeout

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 100: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-100

TCP congestion control FSM details

slow start

congestionavoidance

fastrecovery

timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment timeoutssthresh = cwnd2

cwnd = 1 MSSdupACKcount = 0

retransmit missing segment

Λcwnd gt ssthresh

cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed

new ACKcwnd = cwnd + MSS (MSScwnd)

dupACKcount = 0transmit new segment(s)as allowed

new ACK

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment

dupACKcount == 3

dupACKcount++duplicate ACK

ssthresh= cwnd2cwnd = ssthresh + 3

retransmit missing segment

dupACKcount == 3

timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment

cwnd = cwnd + MSStransmit new segment(s) as allowed

duplicate ACK

cwnd = ssthreshdupACKcount = 0

New ACK

Λ

cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 101: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-101

Popular ldquoflavorsrdquo of TCP

ssthresh

ssthresh

TCPTahoe

TCPReno

Transmissionround

cwnd windowsize(insegments)

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 102: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-102

Summary TCP Congestion Control

when cwnd lt ssthresh sender in slow-startphase window grows exponentially

when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly

when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh

when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 103: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-103

TCP throughput

Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path

For a given cwnd at any point in time

But what is average cwnd as function of L

cwnd =122 MSS

L

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

throughput =mincwnd

RTTrwnd

RTTC

$

amp

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 104: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-104

TCP throughput (more)

throughput =min122 MSS

RTT Lrwnd

RTTC

$ amp

(

Loss rate and RTTlimit throughput dueto AIMD (as cwnd

does not grow quicklyenough before halving

upon loss)

Receive window limitsthroughput due to

flow control (so as tonot overwhelm

receiver)

Bottleneckcapacity obviouslylimits throughput

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 105: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-105

Need better future TCP for ldquolong fat pipesrdquo

Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity

To saturate capacity loss rate formula implies

Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for

long high speed links

10Gbps lt122 MSS

RTT L=122 15008

01 L

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 106: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-106

Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK

TCP connection 1

bottleneckrouter

capacity R

TCP connection 2

TCP Fairness

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 107: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-107

Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally

R

R

equal bandwidth share

Connection 1 throughput

Conn

ecti

on 2

thr

ough

put

congestion avoidance additive increaseloss decrease window by factor of 2

congestion avoidance additive increaseloss decrease window by factor of 2

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 108: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-108

Fairness (more)Fairness and UDP multimedia apps often

do not use TCP do not want rate

throttled by congestioncontrol

instead use UDP pump audiovideo at

constant rate toleratepacket loss

Fairness and parallel TCPconnections

nothing prevents app fromopening parallelconnections between 2hosts

web browsers do this example link of rate R

supporting 9 connections new app asks for 1 TCP gets

rate R10 new app asks for 11 TCPs

gets R2

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo

Page 109: Chapter 3 Transport Layer · 2020. 2. 2. · Transport Layer 3-3 Chapter 3 outline 3.1 Transport-layer services 3.2 Multiplexing and demultiplexing 3.3 Connectionless transport: UDP

Transport Layer 3-109

Chapter 3 Summary principles behind transport

layer services multiplexing

demultiplexing reliable data transfer flow control congestion control

instantiation andimplementation in theInternet UDP TCP

Next leaving the network

ldquoedgerdquo (applicationtransport layers)

into the networkldquocorerdquo