Chapter 5

Peer-to-Peer Protocols
and Data Link Layer
PART I: Peer-to-Peer Protocols
Peer-to-Peer Protocols and Service Models
ARQ Protocols and Reliable Data Transfer
Flow Control
Timing Recovery
TCP Reliable Stream Service & Flow Control
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
PART II: Data Link Controls
Framing
Point-to-Point Protocol
High-Level Data Link Control
Link Sharing Using Statistical Multiplexing
Chapter Overview
z Peer-to-Peer protocols: many protocols involve the
interaction between two peers
z Service Models are discussed & examples given
z Detailed discussion of ARQ provides example of
development of peer-to-peer protocols
z Flow control, TCP reliable stream, and timing recovery
z Data Link Layer
z Framing
z PPP & HDLC protocols
z Statistical multiplexing for link sharing
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Peer-to-Peer Protocols and
Service Models
 Peer-to-Peer Protocols
zzz

zzz
z Peer-to-Peer processes
execute layer-n protocol
to provide service to
n + 1 peer process n + 1 peer process layer-(n+1)

z Layer-(n+1) peer calls

SDU SDU layer-n and passes
PDU Service Data Units
n peer process n peer process (SDUs) for transfer

z Layer-n peers exchange

Protocol Data Units
(PDUs) to effect transfer
n – 1 peer process n – 1 peer process
z Layer-n delivers SDUs to
destination layer-(n+1)
peer
zzz
zzz
Service Models
z The service model specifies the information transfer
service layer-n provides to layer-(n+1)
z The most important distinction is whether the service
is:
z Connection-oriented
z Connectionless
z Service model possible features:
z Arbitrary message size or structure
z Sequencing and Reliability
z Timing, Pacing, and Flow control
z Multiplexing
z Privacy, integrity, and authentication
Connection-Oriented Transfer
Service
z Connection Establishment
z Connection must be established between layer-(n+1) peers
z Layer-n protocol must: Set initial parameters, e.g. sequence
numbers; and Allocate resources, e.g. buffers
z Message transfer phase
z Exchange of SDUs
z Disconnect phase
z Example: TCP, PPP

n + 1 peer process n + 1 peer process

send receive

SDU Layer n connection-oriented service SDU

Connectionless Transfer Service
z No Connection setup, simply send SDU
z Each message send independently
z Must provide all address information per message
z Simple & quick
z Example: UDP, IP

n + 1 peer process n + 1 peer process

send receive

SDU Layer n connectionless service

Message Size and Structure
z What message size and structure will a
service model accept?
z Different services impose restrictions on size &
structure of data it will transfer
z Single bit? Block of bytes? Byte stream?
z Ex: Transfer of voice mail = 1 long message
z Ex: Transfer of voice call = byte stream

1 voice mail= 1 message = entire sequence of speech samples

(a)

1 call = sequence of 1-byte messages

(b)
Segmentation & Blocking
z To accommodate arbitrary message size, a layer may
have to deal with messages that are too long or too
short for its protocol
z Segmentation & Reassembly: a layer breaks long
messages into smaller blocks and reassembles these
at the destination
z Blocking & Unblocking: a layer combines small
messages into bigger blocks prior to transfer
1 long message 2 or more short messages

2 or more blocks 1 block

Reliability & Sequencing
z Reliability: Are messages or information
stream delivered error-free and without loss
or duplication?
z Sequencing: Are messages or information
stream delivered in order?
z ARQ protocols combine error detection,
retransmission, and sequence numbering to
provide reliability & sequencing
z Examples: TCP and HDLC
Pacing and Flow Control
z Messages can be lost if receiving system
does not have sufficient buffering to store
arriving messages
z If destination layer-(n+1) does not retrieve its
information fast enough, destination layer-n
buffers may overflow
z Pacing & Flow Control provide backpressure
mechanisms that control transfer according to
availability of buffers at the destination
z Examples: TCP and HDLC
Timing
z Applications involving voice and video generate
units of information that are related temporally
z Destination application must reconstruct temporal
relation in voice/video units
z Network transfer introduces delay & jitter
z Timing Recovery protocols use timestamps &
sequence numbering to control the delay & jitter in
delivered information
z Examples: RTP & associated protocols in Voice
over IP
Multiplexing
z Multiplexing enables multiple layer-(n+1)
users to share a layer-n service
z A multiplexing tag is required to identify
specific users at the destination
z Examples: UDP, IP
Privacy, Integrity, &
Authentication
z Privacy: ensuring that information transferred
cannot be read by others
z Integrity: ensuring that information is not
altered during transfer
z Authentication: verifying that sender and/or
receiver are who they claim to be
z Security protocols provide these services and
are discussed in Chapter 11
z Examples: IPSec, SSL
End-to-End vs. Hop-by-Hop
z A service feature can be provided by implementing a
protocol
z end-to-end across the network
z across every hop in the network
z Example:
z Perform error control at every hop in the network or only
between the source and destination?
z Perform flow control between every hop in the network or
only between source & destination?
z We next consider the tradeoffs between the two
approaches
 Error control in Data Link Layer
Packets Packets
(a) z Data Link operates
Data link Data link
over wire-like,
layer Frames layer directly-connected
A B systems
Physical Physical
layer layer
z Frames can be
corrupted or lost, but
arrive in order
(b)
1 1 z Data link performs
2 2
error-checking &
12 3 21 12 3 21
Medium retransmission
2 z Ensures error-free
A B 1
packet transfer
1 Physical layer entity between two systems
2 Data link layer entity

3 Network layer entity

 Error Control in Transport Layer
z Transport layer protocol (e.g. TCP) sends segments across
network and performs end-to-end error checking &
retransmission
z Underlying network is assumed to be unreliable

Messages Messages

Transport Segments Transport

layer layer
Network Network Network Network
layer layer layer layer
Data link Data link Data link Data link
layer layer layer layer
End system End system
A Physical Physical Physical Physical
layer layer
B
layer layer

Network
z Segments can experience long delays, can be lost, or
arrive out-of-order because packets can follow different
paths across network
z End-to-end error control protocol more difficult

C
1 2 3 2 1
2 2
1 1

End System
α 1 1 End System
1 2
2 2 β
4 3 21 12 3 2 1 1 2 3 2 1 1 2 3 4
Medium
2
A B 1
Network

3 Network layer entity

4 Transport layer entity

 End-to-End Approach Preferred
Hop-by-hop
Hop-by-hop
cannot ensure
E2E correctness
Data Data Data Data
1 2 3 4 5 Faster recovery
ACK/ ACK/ ACK/ ACK/
NAK NAK NAK NAK

Simple
End-to-end
ACK/NAK
inside the
network

More scalable
1 2 3 4 5 if complexity at
Data Data Data Data the edge
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
ARQ Protocols and Reliable
Data Transfer
Automatic Repeat Request (ARQ)
z Purpose: to ensure a sequence of information
packets is delivered in order and without errors or
duplications despite transmission errors & losses
z We will look at:
z Stop-and-Wait ARQ
z Go-Back N ARQ
z Selective Repeat ARQ
z Basic elements of ARQ:
z Error-detecting code with high error coverage
z ACKs (positive acknowledgments
z NAKs (negative acknowlegments)
z Timeout mechanism
Stop-and-Wait ARQ
Transmit a frame, wait for ACK
Error-free
Packet
packet
Information frame
Transmitter Receiver
Timer set after (Process A) (Process B)
each frame
transmission Control frame

Header
CRC
Information Header
packet CRC
Information frame Control frame: ACKs
Need for Sequence Numbers
(a) Frame 1 lost Time-out
Time
A Frame Frame Frame Frame
0 1 1 2
ACK ACK
B

(b) ACK lost Time-out

Time
A Frame Frame Frame Frame
0 1 1 2
ACK ACK ACK
B

z In cases (a) & (b) the transmitting station A acts the same way
z But in case (b) the receiving station B accepts frame 1 twice
z Question: How is the receiver to know the second frame is also frame 1?
z Answer: Add frame sequence number in header
z Slast is sequence number of most recent transmitted frame
Sequence Numbers
(c) Premature Time-out

Time-out
Time
A Frame
0 Frame Frame Frame
ACK 0 1 2
ACK
B

z The transmitting station A misinterprets duplicate ACKs

z Incorrectly assumes second ACK acknowledges Frame 1
z Question: How is the receiver to know second ACK is for frame 0?
z Answer: Add frame sequence number in ACK header
z Rnext is sequence number of next frame expected by the receiver
z Implicitly acknowledges receipt of all prior frames
1-Bit Sequence Numbering
Suffices
0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1

Slast Rnext
Timer

Transmitter Slast Receiver

A B
Rnext

Global State: Error-free frame 0

(Slast, Rnext) (0,0) (0,1)
arrives at receiver
ACK for
ACK for frame 0
frame 1 arrives at
arrives at transmitter
transmitter Error-free frame 1
arrives at receiver
(1,0) (1,1)
Stop-and-Wait ARQ
Transmitter Receiver
Ready state Always in Ready State
z Await request from higher layer for z Wait for arrival of new frame
packet transfer z When frame arrives, check for errors
z When request arrives, transmit z If no errors detected and sequence
frame with updated Slast and CRC number is correct (Slast=Rnext), then
z Go to Wait State z accept frame,
Wait state z update Rnext,
z Wait for ACK or timer to expire; z send ACK frame with Rnext,
block requests from higher layer z deliver packet to higher layer
z If timeout expires z If no errors detected and wrong
z retransmit frame and reset timer sequence number
z If ACK received: z discard frame
z send ACK frame with Rnext
z If sequence number is incorrect or if
errors detected: ignore ACK z If errors detected
z If sequence number is correct (Rnext z discard frame
= Slast +1): accept frame, go to
Ready state
Applications of Stop-and-Wait
ARQ
z IBM Binary Synchronous Communications
protocol (Bisync): character-oriented data
link control
z Xmodem: modem file transfer protocol
z Trivial File Transfer Protocol (RFC 1350):
simple protocol for file transfer over UDP
Stop-and-Wait Efficiency
Last frame bit ACK
First frame bit enters channel arrives
enters channel Channel idle while transmitter
waits for ACK

A t

B t

First frame bit Receiver

arrives at Last frame bit processes frame
receiver arrives at and
receiver prepares ACK
z 10000 bit frame @ 1 Mbps takes 10 ms to transmit
z If wait for ACK = 1 ms, then efficiency = 10/11= 91%
z If wait for ACK = 20 ms, then efficiency =10/30 = 33%
Stop-and-Wait Model
t0 = total time to transmit 1 frame
A
tproc

B
frame
tprop tproc tack tprop
tf time

t 0 = 2 t prop + 2 t proc + t f + t ack bits/info frame

nf na bits/ACK frame
= 2 t prop + 2 t proc + +
R R

channel transmission rate

 S&W Efficiency on Error-free
channel
Effective transmission rate: bits for header & CRC

number of information bits delivered to destination n f − no

R 0
eff = = ,
total time required to deliver the information bits t0

Transmission efficiency:
Effect of
n f − no no
1− frame overhead
Reff t0 nf
η0 = = = .
R R na 2(t prop + t proc ) R
1+ +
nf nf

Effect of
Effect of
Delay-Bandwidth Product
ACK frame
Example: Impact of Delay-
Bandwidth Product
nf=1250 bytes = 10000 bits, na=no=25 bytes = 200 bits
2xDelayxBW 1 ms 10 ms 100 ms 1 sec
Efficiency
200 km 2000 km 20000 km 200000 km
1 Mbps 103 104 105 106
88% 49% 9% 1%
1 Gbps 106 107 108 109
1% 0.1% 0.01% 0.001%

Stop-and-Wait does not work well for very high speeds

or long propagation delays
S&W Efficiency in Channel with
Errors
z Let 1 – Pf = probability frame arrives w/o errors
z Avg. # of transmissions to first correct arrival is then 1/ (1–
Pf )
z “If 1-in-10 get through without error, then avg. 10 tries to
success”
z Avg. Total Time per frame is then t0/(1 – Pf)
n f − no
no
t0 1−
Reff 1 − Pf nf
η SW = = = (1 − Pf )
R R na 2(t prop + t proc ) R
1+ +
nf nf
Effect of
frame loss
 Example: Impact Bit Error Rate
nf=1250 bytes = 10000 bits, na=no=25 bytes = 200 bits
Find efficiency for random bit errors with p=0, 10-6, 10-5, 10-4
−n f p
1 − Pf = (1 − p ) ≈e
nf
for large n f and small p

1 – Pf 0 10-6 10-5 10-4

Efficiency
1 Mbps 1 0.99 0.905 0.368
& 1 ms 88% 86.6% 79.2% 32.2%

Bit errors impact performance as nfp approach 1

Go-Back-N
z Improve Stop-and-Wait by not waiting!
z Keep channel busy by continuing to send frames
z Allow a window of up to Ws outstanding frames
z Use m-bit sequence numbering
z If ACK for oldest frame arrives before window is
exhausted, we can continue transmitting
z If window is exhausted, pull back and retransmit all
outstanding frames
z Alternative: Use timeout
Go-Back-N ARQ
Go-Back-4: 4 frames are outstanding; so go back 4

fr fr fr fr fr fr fr fr fr fr fr fr fr fr Time
0 1 2 3 4 5 6 3 4 5 6 7 8 9
A

B
A A A out of sequence A A A A A A
C C C frames C C C C C C
K K K K K K K K K
1 2 3 4 5 6 7 8 9

Rnext 0 1 2 3 3 4 5 6 7 8 9

z Frame transmission are pipelined to keep the channel busy

z Frame with errors and subsequent out-of-sequence frames are ignored
z Transmitter is forced to go back when window of 4 is exhausted
Window size long enough to cover round trip time
Time-out expires
Stop-and-Wait ARQ
fr fr fr Time
0 0 1
A

B
Receiver is A
looking for C
K
Rnext=0
1

Go-Back-N ARQ Four frames are outstanding; so go back 4

fr fr fr fr fr fr fr fr fr fr fr
0
Time
1 2 3 0 1 2 3 4 5 6
A

B A A A A A A
Receiver is Out-of- C C C C C C
looking for sequence K K K K K K
Rnext=0 frames 1 2 3 4 5 6
Go-Back-N with Timeout
z Problem with Go-Back-N as presented:
z If frame is lost and source does not have frame to
send, then window will not be exhausted and
recovery will not commence
z Use a timeout with each frame
z When timeout expires, resend all outstanding
frames
 Go-Back-N Transmitter & Receiver
Transmitter Receiver

Send Window
... Receive Window

Frames
transmitted S
last Srecent Slast+Ws-1
and ACKed
Frames
Buffers Rnext
received
oldest un-
Timer Slast
ACKed frame
Timer Slast+1 Receiver will only accept
... a frame that is error-free and
that has sequence number Rnext
Timer
Srecent most recent
transmission
When such frame arrives Rnext is
... incremented by one, so the
Slast+Ws-1 max Seq # receive window slides forward by
allowed one
 Sliding Window Operation
Transmitter

Send Window m-bit Sequence Numbering

...

Frames
transmitted S 0
and ACKed
last Srecent Slast+Ws-1 2m – 1 1

2
Transmitter waits for error-free
ACK frame with sequence
number Slast Slast
send
When such ACK frame arrives,
Slast is incremented by one, and
window i
the send window slides forward i + Ws – 1 i+1
by one
Maximum Allowable Window Size is Ws = 2m-1
M = 22 = 4, Go-Back - 4: Transmitter goes back 4

fr fr fr fr fr fr fr fr Time
A 0 1 2 3 0 1 2 3

A A A A
B C C C C Receiver has Rnext= 0, but it does not
K K K K know whether its ACK for frame 0 was
1 2 3 0
received, so it does not know whether
Rnext 0 1 2 3 0 this is the old frame 0 or a new frame 0

M = 22 = 4, Go-Back-3: Transmitter goes back 3

fr fr fr fr fr fr Time
A 0 1 2 0 1 2

A A A
B C C C
K K K Receiver has Rnext= 3 , so it
1 2 3
rejects the old frame 0
Rnext 0 1 2 3
ACK Piggybacking in Bidirectional GBN
SArecent RA next
Transmitter Receiver
Receiver Transmitter
SBrecent RB next

“A” Receive Window “B” Receive Window

RA next RB next
“A” Send Window “B” Send Window
... ...

SA last SA last+WA s-1 SB last SB last+WB s-1

Buffers Buffers
Timer SA last Timer SB last
SA last+1 SBlast+1
Timer Note: Out-of- Timer
... ...
sequence error-free
Timer SArecent Timer SBrecent
... frames discarded
...
after Rnext examined
Timer S last+W s-1
A A
Timer SB last+WB s-1
Applications of Go-Back-N ARQ
z HDLC (High-Level Data Link Control): bit-
oriented data link control
z V.42 modem: error control over telephone
modem links
Required Timeout & Window Size
Tout
Tproc

Tprop Tf Tf Tprop

z Timeout value should allow for:

z Two propagation times + 1 processing time: 2 Tprop + Tproc
z A frame that begins transmission right before our frame arrives
Tf
z Next frame carries the ACK, Tf
z Ws should be large enough to keep channel busy for Tout
Required Window Size for
Delay-Bandwidth Product

Frame = 1250 bytes =10,000 bits, R = 1 Mbps

2(tprop + tproc) 2 x Delay x BW Window

1 ms 1000 bits 1

10 ms 10,000 bits 2

100 ms 100,000 bits 11

1 second 1,000,000 bits 101

Efficiency of Go-Back-N
z GBN is completely efficient, if Ws large enough to keep
channel busy, and if channel is error-free
z Assume Pf frame loss probability, then time to deliver a frame
is:
z tf if first frame transmission succeeds (1 – Pf
)
z Tf + Wstf /(1-Pf) if the first transmission does not succeed
Pf
Ws t f Ws t f
tGBN = t f (1 − Pf ) + Pf {t f + } = t f + Pf and
1 − Pf 1 − Pf
n f − no no
1−
tGBN nf
ηGBN = = (1 − Pf )
R 1 + (Ws − 1) Pf

Delay-bandwidth product determines Ws

 Example: Impact Bit Error Rate on
GBN
nf=1250 bytes = 10000 bits, na=no=25 bytes = 200 bits
Compare S&W with GBN efficiency for random bit errors with
p = 0, 10-6, 10-5, 10-4 and R = 1 Mbps & 100 ms
1 Mbps x 100 ms = 100000 bits = 10 frames → Use Ws = 11
Efficiency 0 10-6 10-5 10-4

S&W 8.9% 8.8% 8.0% 3.3%

GBN 98% 88.2% 45.4% 4.9%

z Go-Back-N significant improvement over Stop-and-Wait for

large delay-bandwidth product
z Go-Back-N becomes inefficient as error rate increases
Selective Repeat ARQ
z Go-Back-N ARQ inefficient because multiple frames
are resent when errors or losses occur
z Selective Repeat retransmits only an individual frame
z Timeout causes individual corresponding frame to be resent
z NAK causes retransmission of oldest un-acked frame
z Receiver maintains a receive window of sequence
numbers that can be accepted
z Error-free, but out-of-sequence frames with sequence
numbers within the receive window are buffered
z Arrival of frame with Rnext causes window to slide forward by
1 or more
Selective Repeat ARQ

fr fr fr fr fr fr fr fr fr fr fr fr fr fr Time
0 1 2 3 4 5 6 2 7 8 9 10 11 12
A

B
A A N A A A A A A A A A
C C A C C C C C C C C C
K K K K K K K K K K K K
1 2 2 2 2 2 7 8 9 1 1 1
0 1 2
Selective Repeat ARQ
Transmitter Receiver

Send Window Receive Window

...

Frames Frames
transmitted S received Rnext Rnext + Wr-1
last Srecent Slast+ Ws-1
and ACKed

Buffers Buffers
Timer Slast Rnext+ 1
Timer Slast+ 1 Rnext+ 2
...
Timer
Srecent ...
Rnext+ Wr- 1 max Seq #
... accepted
Slast+ Ws - 1
 Send & Receive Windows
Transmitter Receiver

0 0
2m-1 1 2m-1 1

2 2
Rnext
Slast
send receive j
window i window
i
i + Ws – 1 i+1
j + Wr – 1
Moves k forward when ACK Moves forward by 1 or more
arrives with Rnext = Slast + k when frame arrives with
k = 1, …, Ws-1 Seq. # = Rnext
 What size Ws and Wr allowed?
z Example: M=22=4, Ws=3, Wr=3
Frame 0 resent
Send
{0,1,2} {1,2} {2} {.}
Window
fr0 fr1 fr2 fr0
A Time

B ACK1 ACK2 ACK3

Receive
{0,1,2} {1,2,3} {2,3,0} {3,0,1}
Window
Old frame 0 accepted as a
new frame because it falls
in the receive window
 Ws + Wr = 2m is maximum allowed
z Example: M=22=4, Ws=2, Wr=2
Frame 0 resent
Send
{0,1} {1} {.}
Window
fr0 fr1 fr0
A Time

B ACK1 ACK2
Receive
{0,1} {1,2} {2,3}
Window
Old frame 0 rejected because it
falls outside the receive window
 Why Ws + Wr = 2m works
z Transmitter sends frames 0 z Receiver window starts at {0, …, Wr}
to Ws-1; send window empty
z Window slides forward to
z All arrive at receiver {Ws,…,Ws+Wr-1}
z All ACKs lost
z Receiver rejects frame 0 because it
z Transmitter resends frame 0 is outside receive window

0 0
2m-1 1 2m-1 1

Slast Ws +Wr-1 2
2

receive Rnext Ws
send window
window Ws-1
Applications of Selective Repeat
ARQ
z TCP (Transmission Control Protocol):
transport layer protocol uses variation of
selective repeat to provide reliable stream
service
z Service Specific Connection Oriented
Protocol: error control for signaling
messages in ATM networks
Efficiency of Selective Repeat
z Assume Pf frame loss probability, then number of
transmissions required to deliver a frame is:
z tf / (1-Pf)

n f − no
t f /(1 − Pf ) no
η SR = = (1 − )(1 − Pf )
R nf
 Example: Impact Bit Error Rate on
Selective Repeat
nf=1250 bytes = 10000 bits, na=no=25 bytes = 200 bits
Compare S&W, GBN & SR efficiency for random bit errors
with p=0, 10-6, 10-5, 10-4 and R= 1 Mbps & 100 ms

Efficiency 0 10-6 10-5 10-4

S&W 8.9% 8.8% 8.0% 3.3%

GBN 98% 88.2% 45.4% 4.9%

SR 98% 97% 89% 36%

z Selective Repeat outperforms GBN and S&W, but

efficiency drops as error rate increases
 Comparison of ARQ Efficiencies
Assume na and no are negligible relative to nf, and
L = 2(tprop+tproc)R/nf =(Ws-1), then
Selective-Repeat:
no
η SR = (1 − Pf )(1 − ) ≈ (1 − Pf )
nf
Go-Back-N: For Pf≈0, SR & GBN same
1 − Pf 1 − Pf
η GBN = =
1 + (WS − 1) Pf 1 + LPf
For Pf→1, GBN & SW same
Stop-and-Wait:
(1 − Pf ) 1 − Pf
η SW = ≈
n 2 (t + t ) R 1+ L
1+ a + prop proc

nf nf
ARQ Efficiencies
ARQ Efficiency Com parison
Selective
1.5 Repeat
Go Back N 10
Efficiency

0.5 Stop and Wait

100
0 Go Back N 100
-9-9 10
10 -8-8 10
-7-7 10
-6-6 10
-5-5 10
-4-4 10
-3-3 -2
10-2 -1
10-1

- LOG(p)
p Stop and Wait
10

Delay-Bandwidth product = 10, 100

 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Flow Control
Flow Control
buffer fill
Information frame
Transmitter
Receiver

Control frame

z Receiver has limited buffering to store arriving

frames
z Several situations cause buffer overflow
z Mismatch between sending rate & rate at which user can
retrieve data
z Surges in frame arrivals
z Flow control prevents buffer overflow by regulating
rate at which source is allowed to send information
 X ON / X OFF
threshold
Information frame
Transmitter
Receiver

Transmit X OFF Transmit

Time
A

on off on off

B Time

2Tprop

Threshold must activate OFF signal while 2 Tprop R bits still

remain in buffer
 Window Flow Control
Return of permits
tcycle
A
Time

B
Time
z Sliding Window ARQ method with Ws equal to buffer available
z Transmitter can never send more than Ws frames
z ACKs that slide window forward can be viewed as permits to transmit
more
z Can also pace ACKs as shown above
z Return permits (ACKs) at end of cycle regulates transmission rate
z Problems using sliding window for both error & flow control
z Choice of window size
z Interplay between transmission rate & retransmissions
z TCP separates error & flow control
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Timing Recovery
 Timing Recovery for Synchronous
Services
Synchronous source Network output
sends periodic not periodic
information blocks

Network

z Applications that involve voice, audio, or video can generate a

synchronous information stream
z Information carried by equally-spaced fixed-length packets
z Network multiplexing & switching introduces random delays
z Packets experience variable transfer delay
z Jitter (variation in interpacket arrival times) also introduced
z Timing recovery re-establishes the synchronous nature of the stream
 Introduce Playout Buffer
Packet Arrivals Packet Playout
Playout
Buffer

Packet Arrivals

Sequence numbers help order

packets
Packet Playout

Tmax
• Delay first packet by maximum network delay
• All other packets arrive with less delay
• Playout packet uniformly thereafter
 Arrival times
Time
Playout clock must
Send
be synchronized to times Receiver too
transmitter clock Playout slow;
times buffer fills and
overflows

Tplayout time

Receiver too fast Receiver speed

Time buffer starvation Time just right

Many late
packets

Tplayout time Tplayout time

 Clock Recovery
Timestamps inserted in Buffer for information blocks
packet payloads
indicate when info was
Playout
produced
Error command
signal
+ Smoothing Adjust
Add frequency
filter
t4 t3 t2 t1 Recovered
-
Timestamps clock

Counter

z Counter attempts to replicate transmitter clock

z Frequency of counter is adjusted according to arriving timestamps
z Jitter introduced by network causes fluctuations in buffer & in local clock
 Synchronization to a Common
Clock
Transmitter Receiver

M M
fs Network fr

M=#ticks in local clock

In time that net clock Δf=fn-fs=fn-(M/N)fn
does N ticks fn
fn/fs=N/M N ticks N ticks fr=fn-Δf
Network clock

z Clock recovery simple if a common clock is available to transmitter &

receiver
z E.g. SONET network clock; Global Positioning System (GPS)
z Transmitter sends Δf of its frequency & network frequency
z Receiver adjusts network frequency by Δf
z Packet delay jitter can be removed completely
Example: Real-Time Protocol
z RTP (RFC 1889) designed to support real-
time applications such as voice, audio, video
z RTP provides means to carry:
z Type of information source
z Sequence numbers
z Timestamps
z Actual timing recovery must be done by
higher layer protocol
z MPEG2 for video, MP3 for audio
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
TCP Reliable Stream Service &
Flow Control
 TCP Reliable Stream Service
TCP transfers byte
Application Layer stream in order, without Application Layer reads
writes bytes into send
buffer through socket
errors or duplications bytes from receive buffer
through socket
Write 45 bytes
Write 15 bytes Read 40 bytes
Application layer Write 20 bytes Read 40 bytes
Transport layer
Segments

Transmitter Receiver
Send buffer Receive buffer
ACKs
 TCP ARQ Method
• TCP uses Selective Repeat ARQ
• Transfers byte stream without preserving boundaries
• Operates over best effort service of IP
• Packets can arrive with errors or be lost
• Packets can arrive out-of-order
• Packets can arrive after very long delays
• Duplicate segments must be detected & discarded
• Must protect against segments from previous connections
• Sequence Numbers
• Seq. # is number of first byte in segment payload
• Very long Seq. #s (32 bits) to deal with long delays
• Initial sequence numbers negotiated during connection setup
(to deal with very old duplicates)
• Accept segments within a receive window
 Transmitter Receiver

Send Window Receive Window

Slast + Wa-1 Rlast Rlast + WR – 1

... ... ...

octets Rnext Rnew

S S Slast + Ws – 1
transmitted last recent
& ACKed

Slast oldest unacknowledged byte Rlast highest-numbered byte not

Srecent highest-numbered yet read by the application
transmitted byte Rnext next expected byte
Slast+Wa-1 highest-numbered byte Rnew highest numbered byte
that can be transmitted received correctly
Slast+Ws-1 highest-numbered byte Rlast+WR-1 highest-numbered byte
that can be accepted from the that can be accommodated in
application receive buffer
TCP Connections
z TCP Connection
z One connection each way
z Identified uniquely by Send IP Address, Send TCP Port #,
Receive IP Address, Receive TCP Port #
z Connection Setup with Three-Way Handshake
z Three-way exchange to negotiate initial Seq. #’s for
connections in each direction
z Data Transfer
z Exchange segments carrying data
z Graceful Close
z Close each direction separately
 Three Phases of TCP Connection
Host A SYN , S e q _ Host B
no = x

, Ack_ n o = x+1
Three-way n o = y, AC K
SYN, Seq_
Handshake Seq_no = x+1
, ACK, Ack_no
= y+1

Data Transfer
FIN, Seq_no
=w

w+1
ACK, Ack_no =
Graceful
Close Data Transfer =z
FIN, Seq_no
ACK, Ack_no = z+1
1st Handshake: Client-Server
Connection Request

Initial Seq. # from

client to server

SYN bit set indicates request to

establish connection from client to
server
2nd Handshake: ACK from Server

ACK Seq. # =
Init. Seq. # + 1

ACK bit set acknowledges

connection request; Client-
to-Server connection
established
2nd Handshake: Server-Client
Connection Request

Initial Seq. # from

server to client

SYN bit set indicates request to

establish connection from server
to client
3rd Handshake: ACK from Client

ACK Seq. # =
Init. Seq. # + 1

ACK bit set acknowledges

connection request;
Connections in both
directions established
TCP Data Exchange
z Application Layers write bytes into buffers
z TCP sender forms segments
z When bytes exceed threshold or timer expires
z Upon PUSH command from applications
z Consecutive bytes from buffer inserted in payload
z Sequence # & ACK # inserted in header
z Checksum calculated and included in header
z TCP receiver
z Performs selective repeat ARQ functions
z Writes error-free, in-sequence bytes to receive
buffer
Data Transfer: Server-to-Client
Segment

12 bytes of payload
Push set

12 bytes of payload
carries telnet option
negotiation
Graceful Close: Client-to-Server
Connection

Client initiates closing

of its connection to
server
Graceful Close: Client-to-Server
Connection

ACK Seq. # =
Previous Seq. # + 1
Server ACKs request;
client-to-server connection
closed
 Flow Control
z TCP receiver controls rate at which sender transmits to prevent
buffer overflow
z TCP receiver advertises a window size specifying number of
bytes that can be accommodated by receiver
WA = WR – (Rnew – Rlast)
z TCP sender obliged to keep # outstanding bytes below WA
(Srecent - Slast) ≤ WA

Send Window Receive Window

Slast + WA-1
WA
... ... ...

Rlast Rnew Rlast + WR – 1

Slast Srecent Slast + Ws – 1
TCP window flow control
Host A Host B

2 048, No Data t0
= 2000 , Win =
Ack _no
= 1,
Seq_no

Seq_no =
t1 2000, Ack
_no = 1, Win =
1024, Data
= 2000-30
Seq_no = 23
t2 3024, Ack
_no = 1, Win =
1024, Data
= 3024-40
47
28
5 12 , D ata = 1-1 t3
, Win =
, A ck _no = 4048
= 1
Seq_no
Seq_no =
4048, Ack_no =
t4 129, Win
= 1024, D
ata = 404
8-4559
 TCP Retransmission Timeout
z TCP retransmits a segment after timeout period
z Timeout too short: excessive number of retransmissions
z Timeout too long: recovery too slow
z Timeout depends on RTT: time from when segment is sent to
when ACK is received
z Round trip time (RTT) in Internet is highly variable
z Routes vary and can change in mid-connection
z Traffic fluctuates
z TCP uses adaptive estimation of RTT
z Measure RTT each time ACK received: τn

tRTT(new) = α tRTT(old) + (1 – α) τn

z α = 7/8 typical
RTT Variability
z Estimate variance σ2 of RTT variation
z Estimate for timeout:
tout = tRTT + k σRTT
z If RTT highly variable, timeout increase accordingly
z If RTT nearly constant, timeout close to RTT estimate

z Approximate estimation of deviation

dRTT(new) = β dRTT(old) + (1-β) | τn - tRTT |

tout = tRTT + 4 dRTT

 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
PART II: Data Link Controls
Framing
Point-to-Point Protocol
High-Level Data Link Control
Link Sharing Using Statistical Multiplexing
 Data Link Protocols
Packets Packets Data Links Services
z Framing
Data link Data link z Error control
layer Frames layer
A z Flow control
Physical Physical B
z Multiplexing
layer layer
z Link Maintenance
z Security: Authentication &
Encryption

Examples
z Directly connected, wire-like
z PPP
z Losses & errors, but no out-of-
z HDLC
sequence frames
z Ethernet LAN
z Applications: Direct Links;
LANs; Connections across z IEEE 802.11 (Wi Fi) LAN
WANs
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Framing
Framing
transmitted received z Mapping stream of
frames frames physical layer bits into
frames
z Mapping frames into
Framing bit stream
z Frame boundaries can
be determined using:
Character Counts
0110110111
0111110101

z Control Characters
z Flags
z CRC Checks
 Character-Oriented Framing
Data to be sent
A DLE B ETX DLE STX E
After stuffing and framing
DLE STX A DLE DLE B ETX DLE DLE STX E DLE ETX

z Frames consist of integer number of bytes

z Asynchronous transmission systems using ASCII to transmit printable
characters
z Octets with HEX value <20 are nonprintable
z Special 8-bit patterns used as control characters
z STX (start of text) = 0x02; ETX (end of text) = 0x03;
z Byte used to carry non-printable characters in frame
z DLE (data link escape) = 0x10
z DLE STX (DLE ETX) used to indicate beginning (end) of frame
z Insert extra DLE in front of occurrence of DLE STX (DLE ETX) in frame
z All DLEs occur in pairs except at frame boundaries
Framing & Bit Stuffing
HDLC frame

Flag Address Control Information FCS Flag

any number of bits

z Frame delineated by flag character

z HDLC uses bit stuffing to prevent occurrence of flag
01111110 inside the frame
z Transmitter inserts extra 0 after each consecutive
five 1s inside the frame
z Receiver checks for five consecutive 1s
z if next bit = 0, it is removed
z if next two bits are 10, then flag is detected
z If next two bits are 11, then frame has errors
Example: Bit stuffing & de-
stuffing
(a) Data to be sent

0110111111111100
After stuffing and framing

0111111001101111101111100001111110

(b) Data received

01111110000111011111011111011001111110
After destuffing and deframing

*000111011111-11111-110*
 PPP Frame
Flag Address Control Protocol Flag
Information CRC
01111110 1111111 00000011 01111110

integer # of bytes

All stations are to Unnumbered Specifies what kind of packet is contained in the
accept the frame frame payload, e.g., LCP, NCP, IP, OSI CLNP, IPX

z PPP uses similar frame structure as HDLC, except

z Protocol type field
z Payload contains an integer number of bytes
z PPP uses the same flag, but uses byte stuffing
z Problems with PPP byte stuffing
z Size of frame varies unpredictably due to byte insertion
z Malicious users can inflate bandwidth by inserting 7D & 7E
Byte-Stuffing in PPP
z PPP is character-oriented version of HDLC
z Flag is 0x7E (01111110)
z Control escape 0x7D (01111101)
z Any occurrence of flag or control escape inside of frame is
replaced with 0x7D followed by
original octet XORed with 0x20 (00100000)

Data to be sent
41 7D 42 7E 50 70 46

7E 41 7D 5D 42 7D 5E 50 70 46 7E
After stuffing and framing
Generic Framing Procedure
GFP payload area

2 2 2 2 0-60
PLI cHEC Type tHEC GEH GFP payload

Payload Core Payload Type GFP GFP

length header type header extension payload
indicator error error headers
checking checking

z GFP combines frame length indication with CRC

z PLI indicated length of frame, then simply count characters
z cHEC (CRC-16) protects against errors in count field (single-bit
error correction + error detection)
z GFP designed to operate over octet-synchronous physical
layers (e.g. SONET)
z Frame-mapped mode for variable-length payloads: Ethernet
z Transparent mode carries fixed-length payload: storage devices
GFP Synchronization &
Scrambling
z Synchronization in three-states
z Hunt state: examine 4-bytes to see if CRC ok
z If no, move forward by one-byte
z If yes, move to pre-sync state
z Pre-sync state: tentative PLI indicates next frame
z If N successful frame detections, move to sync state
z If no match, go to hunt state
z Sync state: normal state
z Validate PLI/cHEC, extract payload, go to next frame
z Use single-error correction
z Go to hunt state if non-correctable error
z Scrambling
z Payload is scrambled to prevent malicious users from inserting
long strings of 0s which cause SONET equipment to lose bit
clock synchronization (as discussed in line code section)
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Point-to-Point Protocol
PPP: Point-to-Point Protocol
z Data link protocol for point-to-point lines in Internet
z Router-router; dial-up to router
1. Provides Framing and Error Detection
z Character-oriented HDLC-like frame structure
2. Link Control Protocol
z Bringing up, testing, bringing down lines; negotiating
options
z Authentication: key capability in ISP access
3. A family of Network Control Protocols specific to
different network layer protocols
z IP, OSI network layer, IPX (Novell), Appletalk
PPP Applications
PPP used in many point-to-point applications
z Telephone Modem Links 30 kbps
z Packet over SONET 600 Mbps to 10 Gbps
z IP→PPP→SONET

z PPP is also used over shared links such as

Ethernet to provide LCP, NCP, and
authentication features
z PPP over Ethernet (RFC 2516)
z Used over DSL
PPP Frame Format

1 or 2 variable 2 or 4
Flag Address Control Protocol Information FCS Flag
01111110 1111111 00000011 01111110

CRC 16 or
All stations are to
CRC 32
accept the frame

HDLC
Unnumbered frame

• PPP can support multiple network protocols simultaneously

• Specifies what kind of packet is contained in the payload
•e.g. LCP, NCP, IP, OSI CLNP, IPX...
PPP Example
 PPP Phases
1. Carrier Home PC to Internet Service
Dead detected
Provider
7. Carrier 1. PC calls router via modem
dropped 2. PC and router exchange LCP
packets to negotiate PPP
Failed parameters
Terminate Establish
3. Check on identities
2. Options 4. NCP packets exchanged to
6. Done negotiated configure the network layer, e.g.
TCP/IP ( requires IP address
Failed assignment)
5. Open Authenticate 5. Data transport, e.g. send/receive
IP packets
6. NCP used to tear down the
3. Authentication network layer connection (free up
4. NCP completed IP address); LCP used to shut
down data link layer connection
configuration Network
7. Modem hangs up
PPP Authentication
z Password Authentication Protocol
z Initiator must send ID & password
z Authenticator replies with authentication success/fail
z After several attempts, LCP closes link
z Transmitted unencrypted, susceptible to eavesdropping
z Challenge-Handshake Authentication Protocol
(CHAP)
z Initiator & authenticator share a secret key
z Authenticator sends a challenge (random # & ID)
z Initiator computes cryptographic checksum of random # &
ID using the shared secret key
z Authenticator also calculates cryptocgraphic checksum &
compares to response
z Authenticator can reissue challenge during session
Example: PPP connection setup
in dialup modem to ISP

LCP
Setup
PAP
IP NCP
setup
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
High-Level Data Link Control
High-Level Data Link Control
(HDLC)
z Bit-oriented data link control
z Derived from IBM Synchronous Data Link
Control (SDLC)
z Related to Link Access Procedure Balanced
(LAPB)
z LAPD in ISDN
z LAPM in cellular telephone signaling
 NLPDU Network
Network
layer layer
“Packet”

DLSDU DLSAP DLSAP DLSDU

DLPDU
Data link
Data link
layer
layer
“Frame”

Physical Physical
layer layer
HDLC Data Transfer Modes
z Normal Response Mode
z Used in polling multidrop lines
Commands
Primary
Responses

Secondary Secondary Secondary

z Asynchronous Balanced Mode

z Used in full-duplex point-to-point links
Primary Commands Responses Secondary
Secondary Responses Commands Primary

z Mode is selected during connection establishment

HDLC Frame Format

Flag Address Control Information FCS Flag

z Control field gives HDLC its functionality

z Codes in fields have specific meanings and uses
z Flag: delineate frame boundaries
z Address: identify secondary station (1 or more octets)
z In ABM mode, a station can act as primary or secondary so
address changes accordingly
z Control: purpose & functions of frame (1 or 2 octets)
z Information: contains user data; length not standardized, but
implementations impose maximum
z Frame Check Sequence: 16- or 32-bit CRC
 Control Field Format
Information Frame
1 2-4 5 6-8
0 N(S) P/F N(R)

Supervisory Frame

1 0 S S P/F N(R)

Unnumbered Frame

1 1 M M P/F M M M

z S: Supervisory Function Bits z M: Unnumbered Function Bits

z N(R): Receive Sequence Number z P/F: Poll/final bit used in interaction
z N(S): Send Sequence Number between primary and secondary
Information frames
z Each I-frame contains sequence number N(S)
z Positive ACK piggybacked
z N(R)=Sequence number of next frame expected
acknowledges all frames up to and including N(R)-1
z 3 or 7 bit sequence numbering
z Maximum window sizes 7 or 127
z Poll/Final Bit
z NRM: Primary polls station by setting P=1; Secondary
sets F=1 in last I-frame in response
z Primaries and secondaries always interact via paired P/F
bits
Error Detection & Loss Recovery
z Frames lost due to loss-of-synch or receiver buffer
overflow
z Frames may undergo errors in transmission
z CRCs detect errors and such frames are treated as
lost
z Recovery through ACKs, timeouts & retransmission
z Sequence numbering to identify out-of-sequence &
duplicate frames
z HDLC provides for options that implement several
ARQ methods
Supervisory frames
Used for error (ACK, NAK) and flow control (Don’t Send):
z Receive Ready (RR), SS=00
z ACKs frames up to N(R)-1 when piggyback not available
z REJECT (REJ), SS=01
z Negative ACK indicating N(R) is first frame not received
correctly. Transmitter must resend N(R) and later frames
z Receive Not Ready (RNR), SS=10
z ACKs frame N(R)-1 & requests that no more I-frames be sent
z Selective REJECT (SREJ), SS=11
z Negative ACK for N(R) requesting that N(R) be selectively
retransmitted
Unnumbered Frames
z Setting of Modes:
z SABM: Set Asynchronous Balanced Mode
z UA: acknowledges acceptance of mode setting commands
z DISC: terminates logical link connectio
z Information Transfer between stations
z UI: Unnumbered information
z Recovery used when normal error/flow control fails
z FRMR: frame with correct FCS but impossible semantics
z RSET: indicates sending station is resetting sequence
numbers
z XID: exchange station id and characteristics
Connection Establishment &
Release
z Supervisory frames used to establish and release
data link connection
z In HDLC
z Set Asynchronous Balanced Mode (SABM)
z Disconnect (DISC)
z Unnumbered Acknowledgment (UA)

SABM UA
Data UA
transfer DISC
 Example: HDLC using NRM
(polling)Address of secondary
Primary A Secondaries B, C
A polls B B, RR, 0, P N(S) N(R)

B, I, 0, 0
N(R) B sends 3 info
X B, I, 1, 0
frames
B, I, 2, 0,F
A rejects fr1 B, SREJ, 1
A polls C C, RR, 0, P
C, RR, 0, F C nothing to
send
A polls B, B, SREJ, 1,P
requests
selective B, I, 1, 0 B resends fr1
retrans. fr1 B, I, 3, 0 Then fr 3 & 4
B, I, 4, 0, F
A send info fr0 B, I, 0, 5
to B, ACKs up to 4
Time
 Frame Exchange using
Asynchronous Balanced Mode
Combined Station A Combined Station B
B, I, 0, 0 A, I, 0, 0

B, I, 1, 0 A, I, 1, 1 A ACKs fr0
X
B sends 5
frames B, I, 2, 1 A, I, 2, 1

B, I, 3, 2 A rejects
B, REJ, 1
fr1
B, I, 4, 3
A, I, 3, 1
B goes B, I, 1, 3
back to 1
B, I, 2, 4 B, RR, 2 A ACKs fr1
B, I, 3, 4
B, RR, 3 A ACKs fr2
Flow Control
z Flow control is required to prevent transmitter from
overrunning receiver buffers
z Receiver can control flow by delaying
acknowledgement messages
z Receiver can also use supervisory frames to
explicitly control transmitter
z Receive Not Ready (RNR) & Receive Ready (RR)

I3 I4 I5 RNR5 RR6 I6
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
Link Sharing Using Statistical
Multiplexing
 Statistical Multiplexing
z Multiplexing concentrates bursty traffic onto a shared line
z Greater efficiency and lower cost

Header Data payload

B Buffer

Output line
C

Input lines
 Tradeoff Delay for Efficiency
(a) Dedicated lines A1 A2

B1 B2

C1 C2

(b) Shared lines A1 C1 B1 A2 B2 C2

z Dedicated lines involve not waiting for other users, but lines
are used inefficiently when user traffic is bursty
z Shared lines concentrate packets into shared line; packets
buffered (delayed) when line is not immediately available
Multiplexers inherent in Packet
Switches

1 1

2 2

z z
z z
z z

N N

z Packets/frames forwarded to buffer prior to transmission from

switch
z Multiplexing occurs in these buffers
Multiplexer Modeling
Input lines
A
Output line
B Buffer

z Arrivals: What is the packet interarrival pattern?

z Service Time: How long are the packets?
z Service Discipline: What is order of transmission?
z Buffer Discipline: If buffer is full, which packet is dropped?

z Performance Measures:
z Delay Distribution; Packet Loss Probability; Line Utilization
 Delay = Waiting + Service Times

P1 P2 P3 P4 P5
Packet completes
transmission
Service
Packet begins
time
transmission

Waiting
Packet arrives
time
at queue P1 P2 P3 P4
P5
z Packets arrive and wait for service
z Waiting Time: from arrival instant to beginning of service
z Service Time: time to transmit packet
z Delay: total time in system = waiting time + service time
 Fluctuations in Packets in the
System

(a) Dedicated lines A1 A2

B1 B2

C1 C2

(b) Shared line A1 C1 B1 A2 B2 C2

(c) N(t)

Number of
packets in the
system
Packet Lengths & Service Times
z R bits per second transmission rate
z L = # bits in a packet
z X = L/R = time to transmit (“service”) a packet
z Packet lengths are usually variable
z Distribution of lengths → Dist. of service times
z Common models:
z Constant packet length (all the same)
z Exponential distribution
z Internet Measured Distributions fairly constant
ƒ See next chart
Measure Internet Packet
Distribution
z Dominated by TCP
traffic (85%)
z ~40% packets are
minimum-sized 40 byte
packets for TCP ACKs
z ~15% packets are
maximum-sized
Ethernet 1500 frames
z ~15% packets are 552
& 576 byte packets for
TCP implementations
that do not use path
MTU discovery
z Mean=413 bytes
z Stand Dev=509 bytes
z Source: caida.org
M/M/1/K Queueing Model
Exponential service
K – 1 buffer
Poisson Arrivals time with rate μ
rate λ
At most K customers allowed in system

z 1 customer served at a time; up to K – 1 can wait in queue

z Mean service time E[X] = 1/μ
z Key parameter Load: ρ = λ/μ
z When λ << μ (ρ≈0), customers arrive infrequently and usually
find system empty, so delay is low and loss is unlikely
z As λ approaches μ (ρ→1) , customers start bunching up and
delays increase and losses occur more frequently
z When λ > μ (ρ>0) , customers arrive faster than they can be
processed, so most customers find system full and those that
do enter have to wait about K – 1 service times
Poisson Arrivals
z Average Arrival Rate: λ packets per second
z Arrivals are equally-likely to occur at any point in time
z Time between consecutive arrivals is an exponential random
variable with mean 1/ λ
z Number of arrivals in interval of time t is a Poisson random
variable with mean λt

( λ t ) k − λt
P[ k arrivals in t seconds] = e
k!
Exponential Distribution

- t/E [X] - λt
P[ X > t ] = e =e for t > 0 .
Probability density

1-e-λt

P[X<t]
λe-λt

t t
0 0
M/M/1/K Performance Results
(From Appendix A)
Probability of Overflow:
(1 − ρ ) ρ K
Ploss =
1− ρ K +1

Average Total Packet Delay:

ρ ( K + 1) ρ K +1
E[ N ] = −
1− ρ 1− ρ K +1

E[ N ]
E[T ] =
λ (1 − PK )
 M/M/1/10
10
normalized avg delay

9
E[T]/E[X]

8
7
z Maximum 10
5
6 packets allowed in
4
3
system
2
1
z Minimum delay is 1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
service time
1
0.9
load
z Maximum delay is
probability

0.8
10 service times
s probability

0.7
0.6
0.5
0.4
z At 70% load delay
loss

0.3
& loss begin
los

0.2
0.1
0
increasing
0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 2.7 3
lo a d z What if we add
more buffers?
M/M/1 Queue
Exponential service
Infinite buffer
Poisson Arrivals time with rate μ
rate λ
Unlimited number of customers
allowed in system

z Pb=0 since customers are never blocked

z Average Time in system E[T] = E[W] + E[X]
z When λ << μ, customers arrive infrequently and delays are
low
z As λ approaches μ customers start bunching up and
average delays increase
z When λ > μ customers arrive faster than they can be
processed and queue grows without bound (unstable)
Avg. Delay in M/M/1 & M/D/1
10

delay 8

7
avg . delay

6
average

5
no rmalized

4
normalized

M/M/1
3
constant
2
service time
1 M/D/1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.99
lo a d

1⎡ ρ ⎤ ⎡ 1 ⎤1 ⎡ ρ ⎤1 1
E[TM ] = ⎢ =
⎥ ⎢ ⎥ = ⎢ ⎥ + for M/M/1 model.
λ ⎣1 − ρ ⎦ ⎣1 − ρ ⎦ μ ⎣1 − ρ ⎦ μ μ

⎡ ρ ⎤1 ⎡ ρ ⎤1 1
E[T D ] = ⎢1 + ⎥ =⎢ ⎥ + for M/D/1 system.
⎣ 2(1 − ρ ) ⎦ μ ⎣ 2(1 − ρ ) ⎦ μ μ
Effect of Scale
z C = 100,000 bps z C = 10,000,000 bps
z Exp. Dist. with Avg. Packet z Exp. Dist. with Avg. Packet
Length: 10,000 bits Length: 10,000 bits
z Service Time: X=0.1 second z Service Time: X=0.001
second
z Arrival Rate: 7.5 pkts/sec z Arrival Rate: 750 pkts/sec

z Load: ρ=0.75 z Load: ρ=0.75

z Mean Delay: z Mean Delay:

E[T] = 0.1/(1-.75) = 0.4 sec z E[T] = 0.001/(1-.75) =

0.004 sec
Reduction by factor of 100

Aggregation of flows can improve Delay & Loss Performance

Example: Header overhead &
Goodput
z Let R=64 kbps
z Assume IP+TCP header = 40 bytes
z Assume constant packets of total length
z L= 200, 400, 800, 1200 bytes
z Find avg. delay vs. goodput (information transmitted
excluding header overhead)

z Service rate μ = 64000/8L packets/second

z Total load ρ = λ 64000/8L
z Goodput = λ packets/sec x 8(L-40) bits/packet
z Max Goodput = (1-40/L)64000 bps
 Header overhead limits maximum
goodput
0.6
Average Delay (seconds)

0.5 L=1200

0.4
L=800
0.3
L=400
0.2

0.1
L=200
0
0 8000 16000 24000 32000 40000 48000 56000 64000
Goodput (bits/second)
Burst Multiplexing / Speech
Interpolation

Many Fewer
Voice Trunks
Calls

Part of this burst is lost

z Voice active < 40% time
z No buffering, on-the-fly switch bursts to available trunks
z Can handle 2 to 3 times as many calls
z Tradeoff: Trunk Utilization vs. Speech Loss
z Fractional Speech Loss: fraction of active speech lost
z Demand Characteristics
z Talkspurt and Silence Duration Statistics
z Proportion of time speaker active/idle
 Speech Loss vs. Trunks
trunks
10 12 14 16 18 20 22 24
1
speech loss

0.1

0.01

Typical 48
requirement
0.001
24 32 40
# connections
⎛n⎞ k
n

∑
k = m +1
(k − m)⎜ ⎟ p (1 − p ) n − k
⎝k⎠ ⎛n⎞ n!
speech loss = where ⎜ ⎟ = .
np ⎝ ⎠
k k ! ( n − k )!
Effect of Scale
z Larger flows lead to better performance
z Multiplexing Gain = # speakers / # trunks
Trunks required for 1% speech loss
Multiplexing
Speakers Trunks Utilization
Gain
24 13 1.85 0.74
32 16 2.00 0.80
40 20 2.00 0.80
48 23 2.09 0.83
 Packet Speech Multiplexing
A2 A1
Many voice A3
terminals
generating B3 B2 B1
voice packets Buffer B3 C3 A2 D2 C2 B1 C1 D1 A1
C3 C2 C1

D3 D2 D1 Buffer overflow

B2

z Digital speech carried by fixed-length packets

z No packets when speaker silent
z Synchronous packets when speaker active
z Buffer packets & transmit over shared high-speed line
z Tradeoffs: Utilization vs. Delay/Jitter & Loss
 Packet Switching of Voice

1 2 3
Sent t

1 2 3
Received t

z Packetization delay: time for speech samples to fill a packet

z Jitter: variable inter-packet arrivals at destination
z Playback strategies required to compensate for jitter/loss
z Flexible delay inserted to produce fixed end-to-end delay
z Need buffer overflow/underflow countermeasures
z Need clock recovery algorithm
 Chapter 5
Peer-to-Peer Protocols
and Data Link Layer
ARQ Efficiency Calculations
Stop & Wait Performance
1 successful transmission i – 1 unsuccessful transmissions
∞
E [t t o t a l ] = t 0 + ∑ (i − 1)t o ut P[nt = i ]
i =1
∞
=t 0 + ∑ (i − 1)t o ut (1 − Pf )i −1 Pf
i =1

t o u t Pf 1
=t 0 + =t 0 .
1 − Pf 1 − Pf
Efficiency:
n f − no no
1−
E[ttotal ] nf
η SW = = (1 − Pf ) = (1 − Pf )η 0 .
R na 2(t prop + t proc ) R
1+ +
nf nf
Go-Back-N Performance
1 successful transmission i – 1 unsuccessful transmissions
∞
E[ttotal ] = t f + ∑ (i − 1)Ws t f P[nt = i ]
i =1
∞
= t f + Ws t f ∑ (i − 1)(1 − Pf ) i −1 Pf
i =1

Ws t f Pf 1 + (Ws − 1) Pf
= tf + = tf .
1 − Pf 1 − Pf

Efficiency:
n f − no no
1−
E[ttotal ] nf
η GBN = = (1 − Pf ) .
R 1 + (Ws − 1) Pf