1

Chapter 7Data Link Control

Read Chapter 7 & pay attention to the reasons why the Data Link Layer exists (regulates data rate among other functions).

2

Physical Layer

Application

Presentation

Session

transport

Network

Data link

Physical

Application

Presentation

Session

transport

Network

Data link

Physical

Network

Data link

Physical

Source node Destination node

Intermediate node

Signals

Packets

Bits

Frames

3

Data Link LayerProvides for reliable transfer of

information across physical linkIncludes:

transmission of blocks of data (“frames”)

synchronization error control flow control

4

Data Link Layer Functions Frame Synchronization - Create abstraction of “frame-at-

a-time” channel for higher layer (start & end of each frame must be recognizable)

Addressing - Needed when many nodes share transmission link

Flow Control - Control rate of transmission to prevent overflow of receiver's buffers

Error Control - Correct transmission errors (by retransmission) or by error correction schemes

Sequence Control - Receiver must be able to distinguish control information from data

Link Management - Initiation, maintenance, & termination of connections

5

Flow ControlNecessary when data is being sent

faster than it can be processed by receiver

Computer to printer is typical settingCan also be from computer to

computer, when a processing program is limited in capacity

6

Stop-and-Wait Flow ControlSimplest form of Flow ControlSource may not send new frame

until receiver acknowledges the frame already sent

Very inefficient, especially when a single message is broken into separate frames

7

Sliding-Window Flow ControlAllows multiple frames to be in transitReceiver sends acknowledgement with

sequence number of anticipated frameSender maintains list of sequence

numbers it can send, receiver maintains list of sequence numbers it can receive

ACK (acknowledgement) supplemented with RNR (receiver not ready)

8

Error Control ProcessAll transmission media have

potential for introduction of errorsError control process has two

components Error detection Error correction

9

Error DetectionParity CheckCyclic Redundancy Check (CRC)

10

Error CorrectionTwo types of errors

Lost frame Damaged frame

Automatic Repeat reQuest (ARQ) Error detection Positive acknowledgment Retransmission after time-out Negative acknowledgment and retransmission

11

Stop-and-Wait ARQOne frame received and handled at a timeIf frame is damaged, receiver discards it

and sends no acknowledgment Sender uses timer to determine whether or

not to retransmit Sender must keep a copy of transmitted frame

until acknowledgment is receivedIf acknowledgment is damaged, sender

will know it because of numbering

12

Go-Back-N ARQUses sliding-window flow controlWhen receiver detects error, it sends

negative acknowledgment (REJ)Sender must begin transmitting

again from rejected frameTransmitter must keep a copy of all

transmitted frames

13

Flow Control

Technique for controlling data rate so that sender does not over-run receiver

Two approaches exist: 1. Stop-and-wait 2. Sliding-window

14

Stop & Wait Flow Control

1. Sender sends a frame2. Receiver receives frame & acknowledges it3. Sender waits to receive “ack” before

sending next frame (If receiver is not ready to receive another frame it holds back the ack)

* One frame at a time is sent over the transmission line

15

Utilization (Efficiency) of Stop & Wait

Propagation time: time taken for signal to travel from S to R. Thus first bit transmitted at t=0 arrives at R at t = T p = d / V.

16

Utilization (Efficiency) of Stop & WaitTransmission time: time taken to emit all bits of

frame at sender = T t = L / B.

In figure 7.2 Page 197, Transmission Time=1, therefore a = Propagation (Delay) Time

17

Utilization (Efficiency) = U

Note: a=Tp/Tt or Tp=aTtVertical-Time

Sequence Diagram

18

The effect of a on utilization

This shows that for large a (Propagation Time>Transmission Time), the line is under-utilized

19

Utilization (Efficiency) of Stop & Wait

With stop & wait scheme, for high channel utilization, we need a low a (since U= 1/(1+2a))

However, in practice it is desirable to limit frame length L because– error probability increases with L– high average delay with multi-point lines– buffer size limitations

So a more efficient scheme is called for, especially with WAN/satellite communication

20

Utilization (Efficiency) of Stop & Wait Examples

1). WAN using ATM L = 53 bytes = 424 bits, B= 155.52 Mbps, d=1000km T t =

424 / (155.52 x 106 ) = 2.7 x 10-6 seconds assuming optical fiber, T p = 10 6 m / (3 x 108 m/sec) = 0.33 x 10-2 seconds a = (0.33 x 10-2 )/(2.7 x 10-6 ) = 1200 U = 1/2401 = 0.0004

2). LAN d = 0.1 ~ 10 km, B = 10 Mbps, V = 2 x 108 m/sec L = 1000 bits a = 0.005 ~ 0.5 U = 0.5 ~ 0.99

3). Digital trans. vis modem over voice-grade line B = 28.8 kbps, L = 1000 bits, d = 1000 m a = (28.8 kbps *

1000 m) / (2 x 108 m/sec x 1000 bits) = 1.44 x 10-4 U = 1.0, if d = 5000 km, a = (28.8kbps x 5000km) / (2 x 108 x 1000bits) = 0.72 U = 0.4

21

Sliding-Window Flow Control

Pipeline transmission of successive frames Transmit up to “N” frames if necessary without receiving acks. Wait for acks when “N” unacked frames in transit For full duplex transmission each station needs a sending window &

receiving window.

22

Figure 7.4 Example of a sliding-window protocol

23

Utilization: U is a function of a and N

Case 1: N > 1 + 2a : U = 1 Ack for frame 1 reaches Sender before

transmission of Nth frame continuous transmission possible

24

Utilization: U is a function of a and N

Case 2: N < 1 + 2a : U = N / (1 + 2a) Wasted time

between N and 1 + 2a

25

Error Detection

Basic Principle Transmitter: For a given bit stream M,

additional bits (called error-detecting code) are calculated as a function of M and appended to the end of M

Receiver: For each incoming frame, performs the same calculation and compares the two results. A detected error occurs if there is a mismatch

26

Error Detection Two common techniques

Parity checks Cyclic redundancy checks (CRC)

Parity Check One extra “parity” bit is added to each word

Odd parity: bit added so as to make # of 1’s oddEven parity: makes total # of 1’s even

Single parity is very effective with white noise (noise on a line without any active signals on it; e.g., Thermal Noise, see chapter 3), but not very robust with noise bursts (which may extend over whole word duration.)

27

Cyclic Redundancy Checks

Powerful error detection method, easily implemented

Message (M) to be transmitted is appended with extra frame checksum bits (F), so that bit pattern transmitted (T) is perfectly divisible by a special “generator” pattern (P) - (divisor)

At destination, divide received message by the same P. If remainder is nonzero error

28

Cyclic Redundancy ChecksLet

T = (k+n)-bit frame to be transmitted, n<k

M = k-bit message, the first k bits of T F = n-bit FCS, the last n bits of T P = n+1 bits, generator pattern

(predetermined divisor)The concept uses modulo-2 arithmetic

no carries/borrows; add subtract XOR

29

Cyclic Redundancy ChecksMethod

Extend M with n ‘0’s to the right (2 n M)(shift left by n bits

Divide extended message by P to get R (2 n M / P = Q + R/P)

Add R to extended message to form T (T = 2 n M + R)

Transmit T At receiver, divide T by P. Nonzero remainder

means: error QPRRQ

PR

PRQ

PRM

PT n

2

Note: R+R=0 in mod-2 arithmetic

011000

Note: Remainder

R=F=FCS in these

examples

30

Cyclic Redundancy Checks Examples

M = 110011, P = 11001, R = 4 bits 1. Append 4 zeros to M, we get 1100110000

-For each stage of division, if the number of dividend bits equals number of divisor P bits, then Q=1, otherwise Q=0

-Also, see example on page 204

31

Cyclic Redundancy Checks

Exercise: Compute the frame to be transmitted for message 1101011011 using P = 10011

Answer: 11010110111110

32

Cyclic Redundancy Checks

Can view CRC generation in terms of polynomial arithmetic

Any bit pattern polynomial in dummy variable X as shown in the following example: e.g., M = 110011 1·X5 +1·X4 +0·X3

+0·X2 +1·X+1·X0 M(X) = X5 +X4 +X+1

33

Cyclic Redundancy Checks

CRC generation in terms of polynomial Append n ‘0’s Xn M(X) Modulo 2 division Transmit Xn M(X)+R(X) = T(X) At receiver:

34

Cyclic Redundancy Checks

Commonly used polynomials, P(X) CRC-16 = X16 +X15 + X2 +1 CRC-CCITT = X16 +X12 + X5 +1 CRC-32 = X32 +X26 + X23 + X22 +X16 +

X12 + X11 +X10 +X8 + X7 +X12 + X4 + X2 +X+1

35

Can detect1. All single-bit errors2. All double-bit errors, as long as P(X) has a

factor with at least three terms (as long as p has at least three 1s)

3. Any odd number of errors, as long as P(X) contains a factor (X+1)

4. Any burst error for which the length of the burst is less than the length of the FCS

5. Most larger burst errors

36

Why?Error can also be represented by polynomial,

E(X). Tr(X) = T(X) E(X) Error is undetectable if E(X) is divisible by P(X)

P always has at least two terms, Xn , 1 (most & least significant bits equal to 1)

1. Single-bit error: E(X) = Xi (one bit=1) P(X) = X n + … +1 Error is detectable since

E(X) is not divisible by P(X)

T=Transmitted frame

E=Error pattern with 1s in positions where errors happen

Tr=Received frame

37

Why?

2. Double-bit error: E(X) = Xi +Xj = Xi (1+Xk ), k=j-i>0 P(X) does not divide into Xi

P(X) can be chosen which does not divide 1+Xk up to the maximum value of k (i.e., up to the practical frame length). (e.g., X15 + X14 +1 will not divide 1+Xk for any k below 32768)

38

Why?3. No polynomial with an odd # of terms is divisible

by (X+1) Assume E(X) has an odd # of terms and is divisible

by (X+1). Then E(X) = (X+1)Q(X). E(1) = (1+1)Q(1) = 0. However, E(1) cannot be zero since it has an odd # of 1’s

4. A burst error of length r < n can be represented by Xi (Xr-1 + ··· +1). P(X) does not divide into Xi

P(X) which is a polynomial of degree n cannot divide into Xr-1 + ··· +1 since r-1<n.

39

Implementation

Implemented by a circuit consisting of exclusive-or gates and a shift register The shift register contains n bits (length

of FCS) There are up to n exclusive-or gates The presence or absence of a gate

corresponds to the presence or absence of a term in P(X)

40

Implementation

41

Example

Shift Register Circuit for dividing by the polynomial X5 +X4 + X2 +1

Note:R=01110

see pages 206-207

42

Error Control

See page 208

43

Error control techniques

Forward error control: Error recovery by correction at the

receiver [Forward Error Correction (FEC)]

Backward error control: Error recovery by retransmission

[Automatic Repeat Request (ARQ)]

44

ARQBased on

Error detection Positive ack Retransmission after timeout Negative ack. And retransmission

ARQ Stop-and-wait ARQ Continuous ARQ

Go-back-N ARQSelective-reject ARQ

45

Stop & Wait ARQ

Simple and minimum buffer requirement, but inefficient Sender transmits message frame Receiver checks received frame for

errors; sends ACK/NAK Sender waits for ACK/NAK. NAK

retrans; ACK next frame

46

Stop & Wait ARQFrame/ACK could be

lost Uses a timeout mechanism

Possibility of duplication Number frames

Only need a 1-bit frame number alternating 1 and 0 since they are sent one at a time

47

Go-back-N ARQ

If the receiver detects an error on a frame, it sends a NAK for that frame. The receiver will discard all future frames until the frame in error is correctly received. Thus the sender, when it receives a NAK or timeout, must retransmit the frame in error plus all succeeding frames. (Sender must maintain a copy of each unacknowledged frame.)

48

Go-back-N ARQ

49

Selective-reject ARQ

The only frames retransmitted are those that receive a NAK or which timeout

Can save retransmissions, but requires more buffer space and complicated logic

See Figure 7.9b Page 212

50

Maximum window size (with n-bit sequence number)Go-back-N : 2n - 1Selective-reject : 2n-1

51

High-Level Data Link Control (HDLC)

A synchronous data link protocolWidely used, and basis for many other

data link control protocolsConnections can be multipoint or point-to-

pointCan be used in half-duplex or full-duplex

52

Three types of stations

Primary station - (e.g., Mainframe) Controls the operation of the link, issues commands, and receives responses

Secondary station - (e.g., Terminal) Usually only communicates (responds) to a primary station

Combined station - Can be both a primary and a secondary

53

Two link configurations

Unbalanced configuration - One primary and one or more secondary stations

Balanced configurations - Two combined stations

54

Three data transfer modesNormal response mode (NRM) -

Unbalanced configuration Primary station always dictates who sends and receives

Asynchronous balanced mode (ABM) - Two combined stations. Either can initiate transmission

Asynchronous response mode - Unbalanced configuration Secondary station can send at any given time, but only one secondary can be active at a time

55

HDLC Frame StructureThe 1st 3 fields are the header and the last 2 fields are the trailer

0

1

The first one or 2 bits identify the frame type

56

Flag fields

8 bits (01111110)“Bit stuffing” is used for data transparencyBit stuffing: whenever five 1’s are

transmitted, extra zero is inserted

57

Flag fields

An inverted bit could split a frame in two

An inverted bit merges 2 frames

58

HDLC Frame StructureAddress field

8 bits. If needed longer address, the MSB can be set to zero and the address field is then assumed to be 8 bits longer.

All 1’s indicates this is a broadcast frame

59

HDLC Frame Structure

Control field 8 bits. If the 1st bit is 0, then

“information frame”, Otherwise, 10 indicates “supervisory frame” and 11 indicates “unnumbered frame”

Information frametwo 3 bits sequence numbers, P/F bitsequence number can be extended to 7used to send data

60

All of the control field formats contain the Poll/final (P/F) bit. Its use depends on context. Typically, in command frames, it is referred to as the P bit and is set to "1" to solicit (poll) a response frame from the peer HDLC entity. In response frames, it is referred to as the F bit and is set to "1" to indicate the response frame transmitted as a result of a soliciting command.

61

HDLC Frame StructureControl field

Supervisory frame: bits 3 and 4 determine types00: “Receive Ready”. Used to ACK01: “Reject”. Essentially a NACK in Go-back-N

ARQ (Automatic Repeat Request)10: “Receive Not Ready”. Indicates busy

condition11: “Selective Reject”. Request for a single

frame retransmissionBit 5 is P/F. Bits 6,7,8 are sequence number

62

HDLC Frame Structure

Control field Unnumbered frame:

More control functions. Has a variety of purposes, but most often used for establishing the link setup and disconnect.

Sets up the data transfer mode, sequence number size. Also used to reset the link and other miscellaneous stuff.

63

HDLC Commands and responses

64

Examples of Operation