1

EECS F05

EECS 122:Introduction to Computer Networks

Resource Management and QoS

Computer Science DivisionDepartment of Electrical Engineering and Computer Sciences

University of California, BerkeleyBerkeley, CA 94720-1776

2EECS F05

Quality of Service (QoS)

� The Internet’s most contentious subject- Inside vs. Outside the Network (see P&D, pp. 519-520)

� The Internet’s most embarrassing failure- Many papers and proposals, but almost nothing

accomplished

2

3EECS F05

Today’s Lecture

� What “could be”, not what is- Today’s Internet does not have, nor soon will have, a

reasonable QoS solution

� Focus on what one could accomplish with simple (and not-so-simple) mechanisms

- Understand the basic concepts

� Current deployed mechanisms not discussed- Ugly hodge-podge of hacks

4EECS F05

What’s the Problem?

� Internet gives all flows the same “best effort” service- No promises about when or whether packets will be

delivered

� Not all traffic is created equal- Different “owners”, different application requirements- Some applications require service “assurances”

� How can we give traffic different “quality of service”?- Thus begins the problem of QoS

3

5EECS F05

Three Basic Problems

� Control how a link is shared:- Link sharing (discussed by Ion last lecture)

� Give some flows “assured” service- Integrated service (and perhaps differentiated service)

� Give some traffic preferential service- Differentiated service

6EECS F05

A Different Taxonomy

� Giving better service can differ along three dimensions:

- Relative versus absolute- Dropping versus delay

- Flows versus aggregates

� Each choice requires different mechanisms- Intra-Router: Scheduling and dropping decisions- Inter-Router: Signaling protocols

4

7EECS F05

Three Basic Questions

� How does a router service this packet?- Scheduling (various forms of priority and FQ)

- Dropping (fancy versions of RED)

� How does the router know what to do with this packet?

- Bits in packet header or explicit signaling

� How do we control the level of traffic?- Service level agreements (SLAs) or admission control

8EECS F05

Back to Three Basic Problems

� Link sharing (last lecture): not covered today

� Integrated Services

� Differentiated Services

� Nice web site describing DiffServ and IntServ in succinct language: http://www.rhyshaden.com/qos.htm

5

9EECS F05

Integrated Services

� Attempt to integrate service for “real-time”applications into the Internet

� Known as IntServ

� Total, massive, and humiliating failure- 1000s of papers- IETF standards- and STILL no deployment ...

10EECS F05

Key Properties

� All assurances on a per-flow basis

� Traffic can be turned away

� Note:- Co-exists with best-effort service- Similar mechanisms proposed for ATM (Asynchronous

Transfer Mode) but• QoS central in ATM, best-effort an afterthought• Best-effort central in Internet, QoS an afterthought

6

11EECS F05

Example: Video

� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

12EECS F05

Circuit-Switched Networks

� Each packet experiences exactly the same delay

� Packet data is displayed as soon as it arrives

� Signal at receiving end is faithful representation

7

13EECS F05

Internet

� Individual packets experience different delays

� Can’t treat network as a “wire”

� Application must adapt to network service

14EECS F05

Router Effect on Delay

� � � �

� � � � � � � � � � � � �� � � � � � � �

� � �

Delay variationor

Jitter

8

15EECS F05

Router Effects on Traffic

� � � � � � � � �� � �

� � � �

� � �

� � � � � � � � � � � � �

� � � � � �

� � � � � � �

16EECS F05

Network Effects on Traffic

� � � � � � � � �� � �

� � � �

� � �� � � � � �

� � � � � � �

� � � � � � � � � � � � �

� ! " # $ % � � � � & � � ! % � '� � % � ' � � % � � � !# � � � " ( ) � � � � � �* + , - . / , 0 1 2 / 3 0 4 2 . 0 5 4 61 7 3 4 4 1 + 3 8 - . / , 0 1 2 / 3 0

4 2 . 0 5 4 9

� � � � � � �

9

17EECS F05

Network Effects on Traffic

� � � � � � � � �� � �

� � � �

� � �� � � � � �

� � � � � � � � � � � � � � � � � � � �

* + , - . / , 0 1 2 / 3 0 4 2 . 0 5 4 61 7 3 4 4 1 + 3 8 - . / , 0 1 2 / 3 0

4 2 . 0 5 4 9

� � � � � � �

18EECS F05

Network Effects on Traffic

� � � � � � � � �� � �

� � � �

� � �� � � � � �

� � � � � � � � � � � � � � � � � � � �

10

19EECS F05

Network Effect on Delay

� � � �

� � � � � � � � � � � � �� � � � � � � � �

� � �

Delay variationor

Jitter

� � � � � � � � � � �

20EECS F05

Choices

� Play back data upon arrival- Distorted signal

� Buffer data for a while (playback buffer)- Extra delay, less distortion

� Tradeoff depends on application (and use)- Noninteractive: absorb delay, eliminate all distortion- Interactive: absorb only a little delay, eliminate some

distortion

11

21EECS F05

Playback Buffer

� � � � � � � � �� � �

� � � �� � � � � � � �

� � �� � � �

� � � � � � � � � � � � � � �

� � � � �� �

� � � ��

Playback Delay

Play back data a fixed time interval after it was sent

22EECS F05

Playback Point

! " # & " ( �' � � � �

� � � � � � � � � � �

12

23EECS F05

Adaptation

� Can move playback point as delays vary

� Moving playback point:- Increases distortion- But allows lower delays

24EECS F05

Application Taxonomy (Oversimplified and Fanciful)

� Elastic versus “real-time”- Traditional data apps are elastic

- Streaming media are real-time

� RT intolerant versus RT tolerant- Intolerant applications need all data

� Tolerant nonadaptive versus tolerant adaptive- Not clear why any tolerant app couldn’t adapt

� Rate-adaptive versus delay-adaptive (or both)

13

25EECS F05

Key Points

� Some apps don’t need to know maximal delay, just need it to be controlled

- Tolerant, delay-adaptive applications will move playback point to reduce delay

- Can absorb occasional outliers

� Some apps need to know maximal delay- Can’t tolerate loss or distortion- Need to fix playback point and so need a priori

knowledge of delay bound- Bound is typically much worse than actual delays

26EECS F05

Two Service Classes

� Controlled Load- Keep delays under control, but no bound

� Guaranteed Service- Explicit delay bound

14

27EECS F05

Process

� Flow requests service from network- Service request specification (RSpec)

• Controlled load: nothing• Guaranteed: service rate (can calculate delay)

- Traffic specification (TSpec) (next slide)

� Routers decide if they can support request- Admission control

� If so, traffic is classified and scheduled at routers based on per-flow information

28EECS F05

Problem

� How do you describe bursty traffic?

� Network needs some description of traffic

� But video source is bursty (due to coding)- Can’t predict in advance the exact behavior

� Describe “envelope” of traffic: rate and burstiness

� Bits sent between times s and t: A(s,t) � σ + ρ(t-s)

ρ : average rateσ : burstiness

15

29EECS F05

TSpec: The Token Bucket

� � � � � �� � � � � �

ρ � � � � � � � �� � �

� σ

� � � � � � � � � � � �

� � � � � � � � � � � � � �

� � � � ! � " # $% & ' ( � ! ) % � $ ! # ( � �

ρ : average rateσ : burstiness

* + , - . , / 0 1 2 3Bits sent between times s and t: A(s,t) � σ + ρ(t-s)

30EECS F05

Required Elements

4 Reservation Protocol- How service request gets from host to network

4 Admission control algorithm- How network decides if it can accept flow

4 Packet scheduling algorithms (covered last lecture)

- So routers can deliver service

16

31EECS F05

Control Plane vs. Data Plane

4 Control plane: - How information gets to routers

4 Data plane:- What routers do with that information to data packets

Control

Data

32EECS F05

Control Plane: Resource Reservation

SenderReceiver

17

33EECS F05

Control Plane: Resource Reservation

SenderReceiverSender sends Tspec

34EECS F05

Control Plane: Resource Reservation

SenderReceiverPath established

18

35EECS F05

Control Plane: Resource Reservation

SenderReceiver

The receiver signalsreservation request

36EECS F05

Control Plane: Admission Control

SenderReceiverPer-flow state

19

37EECS F05

SenderReceiver

Control Plane: Admission Control

Per-flow state on all routers in path

38EECS F05

Data Plane

SenderReceiver

Per-flow classification on each router

20

39EECS F05

Data Plane

SenderReceiver

Per-flow classification on each router

40EECS F05

Data Plane

SenderReceiverPer-flow scheduling on each router

21

41EECS F05

Resource Reservation Protocol: RSVP

4 Establishes end-to-end reservations over a datagram network

4 Designed for multicast (which will be covered later)

4 Sources: send TSpec4 Receivers: respond with RSpec Network4 Network: responds to reservation requests

42EECS F05

PATH and RESV Messages

4 Sender sends PATH messages - TSPEC: use token bucket- Set up the path state on each router including the

address of previous hop (route pinning)- Collect path information (for guaranteed service)

4 Receiver sends RESV message on the reverse path

- Specify RSpec and TSpec- Sets up the reservation state at each router

22

43EECS F05

The Big Picture

NetworkSender

Receiver

PATH Msg

44EECS F05

The Big Picture

NetworkSender

Receiver

PATH Msg

RESV Msg

23

45EECS F05

Soft State

4 Per session state has a timer associated with it- Path state, reservation state

4 State deleted when timer expires

4 Sender/Receiver periodically refreshes the state, resends PATH/RESV messages, resets timer

4 Advantages:- No need to clean up dangling state after failure- Can tolerate lost signaling packets- Easy to adapt to route changes

46EECS F05

Route Pinning

� Problem: asymmetric routes- You may reserve resources on R

�S3

�S5

�S4

�S1

�S, but data

travels on S�

S1�

S2�

S3�

R !� Solution: use PATH to remember direct path from S to R, i.e.,

perform route pinning

S1S1

S2S2

S3S3

SSRR

S5S5S4S4PATH

RESV

IP routing

24

47EECS F05

Admission Control

4 Parameter-based: worst cast analysis- Guaranteed service

- Low utilization

4 Measurement-based: measure current traffic- Controlled load service- Higher utilization

4 Remember that best-effort service co-exists- No need for IntServ traffic to achieve high utilization

48EECS F05

IntServ Node Architecture

Admission Control

Data InData Out

Con

trol

Pla

neD

ata

Pla

ne

Scheduler

Routing Routing

MessagesRSVP

messages

Classifier

RSVP

Route Lookup

Forwarding Table Per Flow QoS Table