15-744: Computer Networking L-2 Links. L -2; 1-17-01© Srinivasan Seshan, 20012 Links How to make computers talk across a wire How to share the wire How.

15-744: Computer Networking

L-2 Links

L -2; 1-17-01© Srinivasan Seshan, 2001 2

Links

• How to make computers talk across a wire• How to share the wire• How to extend to multiple segments• Assigned reading

• [MB76] ETHERNET: Distributed Packet Switching for Local Area Networks

• [B+88] Measured Capacity of an Ethernet: Myths and Reality


Outline

• Physical media is analog• Modulation – signals to bits

• Bit stream vs. packets• Framing – how to make packets

• Corruption• Error detection & recovery

• Sharing• Media access


Modulation

• Non-Return to Zero (NRZ)• Used by Synchronous Optical Network

(SONET)• 1=high signal, 0=low signal• Long sequence of same bit cause difficulty

• DC bias hard to detect – low and high detected by difference from average voltage

• Clock recovery difficult

• SONET XOR’s bit sequence to ensure frequent transitions


Clock Recovery

• When to sample voltage?• Synchronized sender and receiver clocks• Need easily detectible event at both ends

• Signal transitions help resync sender and receiver

• Need frequent transitions to prevent clock skew


Modulation

• Non-Return to Zero Inverted (NRZI)• 1=inversion of current value, 0=same value• No problem with string of 1’s• NRZ-like problem with string of 0’s


Modulation

• Manchester• Used by Ethernet• 0=low to high transition, 1=high to low transition• Transition for every bit simplifies clock recovery• Not very efficient

• Doubles the number of transitions• Circuitry must run twice as fast


Modulation

• 4b/5b• Used by FDDI• Uses 5bits to encode every 4bits• Encoding ensures no more than 3 consecutive

0’s• Uses NRZI to encode resulting sequence• 16 data values, 3 “special” illegal values, 6

“extra” values, 7 illegal values


Outline






Framing

• Length delimited• Beginning of frame has length• Single corrupt length can cause problems

• Must have start of frame character to resynchronize• Resynchronization can fail if start of frame character

is inside packets as well


Framing

• Byte stuffing• Special start of frame byte (e.g. 0xFF)• Special escape byte value (e.g. 0xFE)• Values actually in text are replaced (e.g. 0xFF

by 0xFEFF and 0xFE by 0xFEFE)• Worst case – can double the size of frame

• Bit stuffing• Special bit sequence (0x01111110)• 0 bit stuffed after any 11111 sequence


Clock-Based Framing

• Used by SONET• Fixed size frames (810 bytes)• Look for start of frame marker that appears

every 810 bytes• Will eventually sync up


Consistent Overhead Byte Stuffing

• Run length encoding applied to byte stuffing• Encoding

• Add implied 0 to end of frame• Each 0 is replaced with (number of bytes to

next 0) + 1• What if no 0 within 255 bytes? – 255 value

indicates 254 bytes followed by no zero• Worst case – no 0’s in packet – 1/254 overhead• Possible optimization to encode series of 0’s


Outline






Error Detection - Checksum

• Used by TCP, UDP, IP, etc..• Ones complement sum of all

words/shorts/bytes in packet• Simple to implement• Relatively weak detection

• Easily tricked by typical loss patterns


Error Detection – CRC

• Polynomial code• Treat packet bits a coefficients of n-bit

polynomial• Choose r+1 bit generator polynomial (well

known – chosen in advance)• Add r bits to packet such that message is

divisible by generator polynomial

• Better loss detection properties than checksums


Error Recovery

• Two forms of error recovery• Forward Error Correction (FEC)• Automatic Repeat Request (ARQ)

• FEC• Use error correcting codes to repair losses

• ARQ• Receiver sends acknowledgement (ACK) when

it receives packet• Sender waits for ACK and timeouts if it does

not arrive within some time period


Stop and Wait

Time

Packet

ACKTim

eou

t

• Simplest ARQ protocol

• Send a packet, stop and wait until acknowledgement arrives

Sender Receiver


Recovering from Error

Packet

ACK

Tim

eou

t

Packet

ACK

Tim

eou

t

Packet

Tim

eou

t

Packet

ACKT

ime

out

Time

Packet

ACK

Tim

eou

t

Packet

ACK

Tim

eou

t

ACK lost Packet lost Early timeout


Problems with Stop and Wait

• How to recognize a duplicate

• Performance• Can only send one packet per round trip


How to Recognize Resends?

• Use sequence numbers• both packets and acks

• Sequence # in packet is finite -- how big should it be? • For stop and wait?

• One bit – won’t send seq #1 until received ACK for seq #0

Pkt 0

ACK 0

Pkt 0

ACK 1

Pkt 1ACK 0


How to Keep the Pipe Full?

• Send multiple packets without waiting for first to be acked• Number of pkts in flight = window

• How large a window is needed• Round trip delay * bandwidth =

capacity of pipe• Reliable, unordered delivery

• Several parallel stop & waits• Send new packet after each ack• Sender keeps list of unack’ed

packets; resends after timeout• Receiver same as stop&wait


Sliding Window

• Reliable, ordered delivery• Receiver has to hold onto a packet until all

prior packets have arrived• Sender must prevent buffer overflow at

receiver• Circular buffer at sender and receiver

• Packets in transit <= buffer size • Advance when sender and receiver agree

packets at beginning have been received


ReceiverReceiverSenderSender

Sender/Receiver State

… …

Sent & Acked Sent Not Acked

OK to Send Not Usable

… …

Max acceptable

Receiver window

Max ACK received Next seqnum

Received & Acked Acceptable Packet

Not Usable

Sender window

Next expected


Window Sliding – Common Case

• On reception of new ACK (i.e. ACK for something that was not acked earlier• Increase sequence of max ACK received• Send next packet

• On reception of new in-order data packet (next expected)• Hand packet to application• Send cumulative ACK – acknowledges

reception of all packets up to sequence number• Increase sequence of max acceptable packet


Loss Recovery

• On reception of out-of-order packet• Send nothing (wait for source to timeout)• Cumulative ACK (helps source identify loss)

• Timeout (Go Back N recovery)• Set timer upon transmission of packet• Retransmit max ACK received sequence + 1• Restart from max ACK received sequence + 1

• Performance during loss recovery• No longer have an entire window in transit• Can have much more clever loss recovery

• Covered in TCP lectures


Sequence Numbers

• How large do sequence numbers need to be?• Must be able to detect wrap-around• Depends on sender/receiver window size

• E.g.• Max seq = 7, send win=recv win=7• If pkts 0..6 are sent succesfully and all acks lost

• Receiver expects 7,0..5, sender retransmits old 0..6

• Max sequence must be >= send window + recv window


Outline






Broadcast Network Arbitration

• Like sharing a single cpu must share a link • Give everyone a fixed time/freq slot?

• Ok for fixed bandwidth (e.g., voice)• What if traffic is bursty?

• Centralized arbiter• Ex: cell phone base station• Single point of failure

• Distributed arbitration• Aloha/Ethernet


Multiplexing/Media Access/Sharing

• FDMA• Different freq

• CDMA (covered in wireless lecture)• Different codes • Direct sequence• Frequency hopping

• TDMA• Different times• Synchronous TDMA• CSMA, etc


Ethernet

• First practical local area network, built at Xerox PARC in 70’s

• Carrier sense• Check to see if active transmission

• Collision detect• Sender checks for collision; wait and retry

• Adaptive randomized wait to avoid collisions


Ethernet MAC Protocol

Packet?

Sense Carrier

Discard Packet

Send Detect Collision

b=CalcBackoff(); wait(b);

attempts++;

No

Yes

attempts < 16

attempts == 16


Ethernet Backoff Calculation

• If deterministic delay after collision, collision will occur again in lockstep

• If random delay with fixed mean• Few senders needless waiting• Too many senders too many collisions

• Exponentially increasing random delay• Infer senders from # of collisions• More senders increase wait time


Minimum Packet Size

• What if two people sent really small packets• How do you find

collision?• None in paper – but

is one in reality


Ethernet Collision Detect

• Min packet length > 2x max prop delay• If A, B are at opposite sides of link, and B starts

one link prop delay after A

• Jam network for 32-48 bits after collision, then stop sending• Ensures that everyone notices collision


End to End Delay

• 1Km, c in cable = 60% * c in vacuum = 1.8 x 10^8 m/s• 1000/1.8 x 10^8 ~= 5 x 10^-6 = 5us• 5us * 3Mbps = 15bits in flight!

• Modern 10Mb Ethernet {• 2.5km, 10Mbps • ~= 12.5us delay• +introduced repeaters (max 5 segments)• worst case – 51.2us round trip time!


Minimum Packet Size

• Slot time = 51.2us = 512bits in flight• After this amount, sender is guaranteed sole access to

link• 51.2us = slot time for backoff

• What about scaling? 100Mbit, 1Gbit...• Make network smaller?

• Solution for 100BaseT• Make min pkt size larger?

• 512bits @ 1Gbps = 512ns• 512ns * 1.8 * 10^8 = 92meters• Gigabit ethernet uses collision extension for small pkts


Ethernet Packet Format

• Preamble – 8 byte• 101010…1011

• Src/Dst Address – 6 bytes• Globally unique• Allocated to manufacturers

• Type – 2 bytes• Data – 46 to 1500 bytes• CRC – 4 bytes

Preamble Dst. Addr. TypeSrc. Addr. Data CRC


Interaction with Layering

• TYPE field• Identifies upper layer protocol• Enables the multiplexing of protocols

• ARP• Broadcast search for IP address• Destination responds with appropriate 48-bit

Ethernet address


Simple Analysis of Efficiency

• If pkt size/link speed = P/C = slot time = T• Minimum packet size• E = 1/(1+W) = 1/ (1 + (1-A)/A) = A• A = (1-(1/Q))^(Q-1)• If Q >>1 the E = A = 1/e = well known Ethernet limit

• Assumptions • Single pkt size• Always have pkt to send• Master slot clock, all transmissions start a slot start• Q is estimated well by exponential backoff


Ethernet Problems

• Ethernet unstable at high loads• Peak utilization = 1/e = 37%

• Peak throughput worse with• More hosts – more collisions needed to identify

single sender• Smaller packet sizes – more frequent

arbitration• Longer links – collisions take longer to observe,

more wasted bandwidth


Source of Myths

• Assumptions in various analyses• Static parameters

• Maximum propagation time and slot time• Maximum packet size• 1-persistence

• Workload parameters• Packet length distribution• Number of hosts• Length of cable and placement of hosts• Load

• Evaluated under unreasonable high load


Measurement of Real Ethernet

• Evaluate performance in some typical scenarios• Scenario 1

• Topology: 4 clusters of 6 hosts – similar to office configuration

• Fixed pkt size• Throughput decreases with number of hosts &

increases with pkt size – as expected• Fairness improves with number of hosts – capture

effects less likely• Only linear increase in delay with number of hosts -

unexpected


Measurement of Real Ethernet

• Scenario 2• Topology: 23 hosts on short net• Load: fixed pkt size• Improvement in bit rate over scenario 1

• Scenario 3• Topology: 4 clusters• Load: bimodal pkt size• 7/1 ratio of small to large pkts is sufficient to

greatly improve total bit rate


How to Improve Performance

• No long cables• Fewer hosts per cable• Use large packets• Don't mix real-time w/ bulk-data if possible

• Can’t provide good efficiency/throughput and good latency


Ethernet Packet Traces

• Ethernet traffic is “self-similar” (fractal)• Bursty at every time scale (msecs to months)

• Implication?• On average, low load• Occasional peaks


Token Rings

• Packets broadcast around ring• Token “right to send” rotates around ring

• Fair, real-time bandwidth allocation• Every host holds token for limited time• Higher latency when only one sender

• Higher bandwidth• Point to point links electrically simpler than bus


Why Did Ethernet Win?

• Failure modes• Token rings – network unusable• Ethernet – node detached

• Good performance in common case• Volume lower cost higher volume ….• Adaptable

• To higher bandwidths (vs. FDDI)• To switching (vs. ATM)

• Completely distributed, easy to maintain/administer• Easy incremental deployment• Cheap cabling, etc


LAN Switching

• Extend reach of a single shared medium• Connect two or more “segments” by

copying data frames between them

LAN 1 LAN 2


Switched Network Advantages

• Higher link bandwidth• Point to point electrically simpler than bus

• Much greater aggregate bandwidth• Separate segments can send at once

• Improved fault tolerance• Redundant paths

• Challenge• Learning which packets to copy across links• Avoiding forwarding loops


Learning Bridges

• Basic idea: build cache of which nodes are downstream of which ports

• Monitor source address on all packets bridge forwards

• Insert reverse path to source into cache• What to do with unknown sources?

• Flood network


Bridges with Loops

B1 B2

LAN 1

LAN 2

Port 21

Port 22

Port 11

Port 12

S

Node

S

Port

11

Cache at B1

Node

S

Port

21

Cache at B2

S D Type DataPacket from S on LAN 1


Loop Solution: Spanning Tree

• Only want a single spanning tree• First step agree on a root• Lowest node ID

• Each bridge announces its current root and distance to root

• Lowest distance become responsible for link

• Ties go to lower bridge ID


Next Lecture: Layer Functionality

• How to determine split of functionality• Across protocol layers• Across network nodes

• Assigned Reading• [SRC84] End-to-end Arguments in System

Design• [Cla88] Design Philosophy of the DARPA

Internet Protocols• [CT90] Architectural Consideration for a New

Generation of Protocols

15-744: Computer Networking L-2 Links. L -2; 1-17-01© Srinivasan Seshan, 20012 Links How to make computers talk across a wire How to share the wire How.

Documents

srinivasan seshan

sequence slide

links slide

reality slide

fast slide

clock skew slide

size of frame

frame character