wolf8

7/27/2019 wolf8

1/62

2000 MorganKaufman Overheads for Computers asComponents

Networking for Embedded

Systems

Why we use networks.

Network abstractions.

Example networks.

7/27/2019 wolf8

2/62

2000 MorganKaufman Overheads for Computers asComponents

Network elements

PEPE

PE

network

communication link

distributed computing platform:

PEs may be CPUs or ASICs.

7/27/2019 wolf8

3/62

2000 MorganKaufman Overheads forComputers as

Components

Networks in embedded

systems

PE

PE sensor

PE actuator

initial processing

more processing

7/27/2019 wolf8

4/62

2000 Morgan

Kaufman

Overheads for Computers as

Components

Why distributed?

Higher performance at lower cost.

Physically distributed activities---time

constants may not allow transmission tocentral site.

Improved debugging---use one CPU in

network to debug others.May buy subsystems that have embedded

processors.

7/27/2019 wolf8

5/62

2000 Morgan

Kaufman


Components

Network abstractions

International Standards Organization(ISO) developed the Open Systems

Interconnection(OSI) model to describenetworks:

7-layer model.

Provides a standard way to classifynetwork components and operations.

7/27/2019 wolf8

6/62

2000 Morgan

Kaufman


Components

OSI model

physical mechanical, electrical

data link reliable data transport

network end-to-end service

transport connections

presentation data format

session application dialog control

application end-use interface

7/27/2019 wolf8

7/62

2000 Morgan

Kaufman


Components

OSI layers

Physical: connectors, bit formats, etc.

Data link: error detection and control

across a single link (single hop).Network: end-to-end multi-hop data

communication.

Transport: provides connections; mayoptimize network resources.

7/27/2019 wolf8

8/62

2000 Morgan

Kaufman


Components

OSI layers, contd.

Session: services for end-userapplications: data grouping,

checkpointing, etc.Presentation: data formats,

transformation services.

Application: interface between networkand end-user programs.

7/27/2019 wolf8

9/62

2000 Morgan

Kaufman


Components

Hardware architectures

Many different types of networks:

topology;

scheduling of communication;routing.

7/27/2019 wolf8

10/62

2000 Morgan

Kaufman


Components

Point-to-point networks

One source, one or more destinations, nodata switching (serial port):

PE 1 PE 2 PE 3

link 1 link 2

7/27/2019 wolf8

11/62

2000 Morgan

Kaufman


Components

Bus networks

Common physical connection:

PE 1 PE 2 PE 3 PE 4

header address data ECC packet format

7/27/2019 wolf8

12/62

2000 Morgan

Kaufman


Components

Bus arbitration

Fixed: Same order of resolution everytime.

Fair: every PE has same access over longperiods.

round-robin: rotate top priority among Pes.

A,B,C A,B,C

fixed

round-robin

A B C A B C

A B C AB C

7/27/2019 wolf8

13/62

2000 Morgan

Kaufman


Components

Crossbar

in1 in2 in3 in4

out1

out2

out3

out4

7/27/2019 wolf8

14/62

2000 Morgan

Kaufman


Components

Crossbar characteristics

Non-blocking.

Can handle arbitrary multi-cast

combinations.Size proportional to n2.

7/27/2019 wolf8

15/62

2000 Morgan

Kaufman


Components

Multi-stage networks

Use several stages of switching elements.

Often blocking.

Often smaller than crossbar.

7/27/2019 wolf8

16/62

2000 Morgan

Kaufman


Components

Message-based

programming

Transport layer provides message-basedprogramming interface:

send_msg(adrs,data1);Data must be broken into packets at

source, reassembled at destination.

Data-push programming: make thingshappen in network based on datatransfers.

7/27/2019 wolf8

17/62

2000 Morgan

Kaufman


Components

I2C bus

Designed for low-cost, medium data rateapplications.

Characteristics:serial;

multiple-master;

fixed-priority arbitration.Several microcontrollers come with built-

in I2C controllers.

7/27/2019 wolf8

18/62

2000 Morgan

Kaufman


Components

I2C physical layer

master 1 master 1

slave 1 slave 1

SCL

SDL data line

clock line

7/27/2019 wolf8

19/62

2000 Morgan

Kaufman


Components

I2C data format

SCL

SDL

...

MSBstart

...

ack

...

7/27/2019 wolf8

20/62

2000 Morgan

Kaufman


Components

I2C electrical interface

SDL

+Open collector interface:

SCL

+

7/27/2019 wolf8

21/62

2000 Morgan

Kaufman


Components

I2C signaling

Sender pulls down bus for 0.

Sender listens to bus---if it tried to send a

1 and heard a 0, someone else issimultaneously transmitting.

Transmissions occur in 8-bit bytes.

7/27/2019 wolf8

22/62

2000 Morgan

Kaufman


Components

I2C data link layer

Every device has an address (7 bits instandard, 10 bits in extension).

Bit 8 of address signals read or write.General call address allows broadcast.

7/27/2019 wolf8

23/62

2000 Morgan

Kaufman


Components

I2C bus arbitration

Sender listens while sending address.

When sender hears a conflict, if its

address is higher, it stops signaling.Low-priority senders relinquish control

early enough in clock cycle to allow bit to

be transmitted reliably.

7/27/2019 wolf8

24/62

2000 Morgan

Kaufman


Components

I2C transmissions

multi-byte write

read from slave

write, then read

S adrs 0 data data P

S adrs 1 data P

S adrs 0 data S adrs 1 data P

7/27/2019 wolf8

25/62

2000 Morgan

Kaufman


Components

Multiprocessor networks

Multiple DSPs are often connected byhigh-speed networks for signal

processing:

DSP DSP

DSP DSP

7/27/2019 wolf8

26/62

2000 Morgan

Kaufman


Components

SHARC link ports

Six per CPU.

Four bits per link port.

Packets have 32- or 48-bit payload.

Can be controlled by DMA.

Are half-duplex---must be turned around

by program.

7/27/2019 wolf8

27/62

2000 Morgan

Kaufman


Components

Ethernet

Dominant non-telephone LAN.

Versions: 10 Mb/s, 100 Mb/s, 1 Gb/s

Goal: reliable communication over anunreliable medium.

7/27/2019 wolf8

28/62

2000 Morgan

Kaufman


Components

Ethernet topology

Bus-based system, several possiblephysical layers:

A B C

7/27/2019 wolf8

29/62

2000 Morgan

Kaufman


Components

CSMA/CD

Carrier sense multiple access with collisiondetection:

sense collisions;exponentially back off in time;

retransmit.

7/27/2019 wolf8

30/62

2000 Morgan

Kaufman


Components

Exponential back-off times

time

7/27/2019 wolf8

31/62

2000 Morgan

Kaufman


Components

Ethernet packet format

preamblestart

frame

source

adrs

dest

adrs

data

payloadlength padding CRC

7/27/2019 wolf8

32/62

2000 Morgan

Kaufman


Components

Ethernet performance

Quality-of-service tends to non-linearlydecrease at high load levels.

Cant guarantee real-time deadlines.However, may provide very good serviceat proper load levels.

7/27/2019 wolf8

33/62

2000 Morgan

Kaufman


Components

Internet Protocol

Internet Protocol(IP) is basis for Internet.

Provides an internetworkingstandard:

between two Ethernets, Ethernet andtoken ring, etc.

Higher-level services are built on top of

IP.

7/27/2019 wolf8

34/62

2000 Morgan

Kaufman


Components

IP in communication

physical

data link

network

transport

presentation

application

session

physical

data link

network

transport

presentation

application

session

physical

data link

network

node A router node B

IP

7/27/2019 wolf8

35/62

2000 Morgan

Kaufman


Components

IP packet

Includes:

version, service type, length

time to live, protocolsource and destination address

data payload

Maximum data payload is 65,535 bytes.

7/27/2019 wolf8

36/62

2000 Morgan

Kaufman


Components

IP addresses

32 bits in early IP, 128 bits in IPv6.

Typically written in form xxx.xx.xx.xx.

Names (foo.baz.com) translated to IPaddress by domain name server(DNS).

7/27/2019 wolf8

37/62

2000 Morgan

Kaufman


Components

Internet routing

Best effort routing:

doesnt guarantee data delivery at IP layer.

Routing can vary:session to session;

packet to packet.

Hi h l l I t t

7/27/2019 wolf8

38/62

2000 Morgan

Kaufman


Components

Higher-level Internet

services

Transmission Control Protocol(TCP)provides connection-oriented service.

Quality-of-service (QoS) guaranteedservices are under development.

7/27/2019 wolf8

39/62

2000 Morgan

Kaufman


Components

The Internet service stack

IP

UDP

SNMP

TCPUser

Datagram

Protocol

FTP HTTP SMTP telnet

7/27/2019 wolf8

40/62

2000 Morgan

Kaufman


Components

Networks

Network-based design.

Communication analysis.

System performance analysis.Internet-enabled systems.

7/27/2019 wolf8

41/62

2000 Morgan

Kaufman


Components

Communication analysis

First, understand delay for singlemessage.

Delay for multiple messages depends on:network protocol;

devices on network.

7/27/2019 wolf8

42/62

2000 Morgan

Kaufman


Components

Message delay

Assume:

single message;

no contention.Delay:

tm= tx+ tn+ tr

= xmtr overhead + network xmit time +rcvr overhead

Example: I2C message

7/27/2019 wolf8

43/62

2000 Morgan

Kaufman


Components

Example: I2C message

delay

Network transmission time dominates.

Assume 100 kbits/sec, one 8-bit byte.

Number of bits in packet:npacket= start + address + data + stop

= 1 + 8 + 8 + 1 = 18 bits

Time required to transmit: 1.8 x 10

-4

sec.20 instructions on 8 MHz controller adds

2.5 x 10-6delay on xmtr, rcvr.

7/27/2019 wolf8

44/62

2000 Morgan

Kaufman


Components

Multiple messages

If messages can interfere with each other,analysis is more complex.

Model total message delay:ty= td+ tm

= wait time for network + message delay

7/27/2019 wolf8

45/62

2000 Morgan

Kaufman


Components

Arbitration and delay

Fixed-priority arbitration introducesunbounded delay for all but highest-

priority device.Unless higher-priority devices are known to

have limited rates that allow lower devices totransmit.

Round-robin arbitration introducesbounded delay proportional to N.

Example: adjusting

7/27/2019 wolf8

46/62

2000 Morgan

Kaufman


Components

Example: adjusting

messages to reduce delay

Task graph: Network:

P1 P2

P3

d1 d2

M1 M2 M3

allocation3

4

3execution time

Transmission time = 4

7/27/2019 wolf8

47/62

2000 Morgan

Kaufman


Components

Initial schedule

time

M1

M2

M3

network

0 20105 15

P1

P2

d1 d1

P3

Time = 15

7/27/2019 wolf8

48/62

2000 Morgan

Kaufman


Components

New design

Modify P3:

reads one packet of d1, one packet of d2

computes partial resultcontinues to next packet

7/27/2019 wolf8

49/62

2000 Morgan

Kaufman


Components

New schedule

time

M1

M2

M3

network

0 20105 15

P1

P2

d1

P3

d2d1

P3

d2d1

P3

d2d1

P3

d2

Time = 12

7/27/2019 wolf8

50/62

2000 Morgan

Kaufman


Components

Further complications

Acknowledgment time.

Transmission errors.

Priority inversion in

7/27/2019 wolf8

51/62

2000 Morgan

Kaufman


Components

Priority inversion in

networks

In many networks, a packet cannot beinterrupted.

Result is priority inversion:low-priority message holds up higher-priority

message.

Doesnt cause deadlock, but can slowdown important communications.

7/27/2019 wolf8

52/62

2000 Morgan

Kaufman


Components

Multihop networks

In multihop networks, one node receivesmessage, then retransmits to destination

(or intermediate).

A B C

hop 1 hop 2

Network 1 Network 2

System performance

7/27/2019 wolf8

53/62

2000 Morgan

Kaufman


Components

System performance

analysis

System analysis is difficult in general.

multiprocessor performance analysis is hard;

communication performance analysis is hard.Simple example: uncertainty in P1 finish

time -> uncertainty in P2 start time.

P1 P2

7/27/2019 wolf8

54/62

2000 Morgan

Kaufman


Components

Analysis challenges

P2 and P3 can delayeach other, eventhough they are in

separate tasks.

Delays in P1propagate to P2, then

P3, then to P4.

P2

P3

P1

P4

7/27/2019 wolf8

55/62

2000 Morgan

Kaufman


Components

Lower bounds on system

Computationalrequirements:

sum up process

requirements overleast-common multipleof periods, averageover one period.

Communicationrequirements:

Count all

transmissions in oneperiod.

7/27/2019 wolf8

56/62

2000 Morgan

Kaufman


Components

Hardware platform design

Need to choose:

number and types of PEs;

number and types of networks.Evaluate a platform by allocating

processes, scheduling processes and

communication.

7/27/2019 wolf8

57/62

2000 Morgan

Kaufman


Components

I/O-intensive systems

Start with I/O devices, then considercomputation:

inventory required devices;identify critical deadlines;

chooses devices that can share PEs;

analyze communication times;

choose PEs to go with devices.

Computation-intensive

7/27/2019 wolf8

58/62

2000 Morgan

Kaufman


Components

Computation-intensive

systems

Start with shortest-deadline tasks:

Put shortest-deadline tasks on separate PEs.

Check for interference on criticalcommunications.

Allocate low-priority tasks to common PEswherever possible.

Balance loads wherever possible.

Internet-enabled

7/27/2019 wolf8

59/62

2000 Morgan

Kaufman


Components

Internet-enabled

embedded system

Internet-enabled embedded system: anyembedded system that includes an

Internet interface (e.g., refrigerator).Internet appliance: embedded system

designed for a particular Internet task

(e.g. email).

7/27/2019 wolf8

60/62

2000 Morgan

Kaufman


Components

Examples

Laser printer.

Personal digital assistant (PDA).

Home automation system.

7/27/2019 wolf8

61/62

2000 Morgan

Kaufman


Components

Example: Javacam

Hardware platform:

parallel-port camera;

National Semi NS486SXF;1.5 Mbytes memory.

Uses memory-efficient Java Nanokernel.

7/27/2019 wolf8

62/62

2000 Morgan Overheads for Computers as

Javacam architecture

Web browser

QuickCam

486

Java nanokernel

Java VM

HTTPQuickcam

server

QuickCam

applet

wolf8

Documents

Transcript of wolf8