Top Banner
Chap 4 Multiaccess Communication (Part 2) Ling-Jyh Chen
50
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: 4_MA_2

Chap 4 Multiaccess Communication(Part 2)

Ling-Jyh Chen

Page 2: 4_MA_2

2

Classification of Multiple Access Protocols

Multiple access protocols

Contention-based Conflict-free

Random access Collision resolution

FDMA,

TDMA,

CDMA,

Token Bus,

etc

ALOHA,

CSMA,

BTMA,

etc

TREE,

WINDOW, etc

BTMA: Busy Tone Multiple Access

Page 3: 4_MA_2

3

Contention Protocols

ALOHA Developed in the 1970s for a packet radio network

by Hawaii University. Whenever a station has a data, it transmits.

Sender finds out whether transmission was successful or experienced a collision by listening to the broadcast from the destination station. Sender retransmits after some random time if there is a collision.

Slotted ALOHA Improvement: Time is slotted and a packet can

only be transmitted at the beginning of one slot. Thus, it can reduce the collision duration.

Page 4: 4_MA_2

4

Slotted ALOHA

1 2&3 2Time

Collision

Retransmission Retransmission

3

Slot

Node 1 Packet

Nodes 2 & 3 Packets

Page 5: 4_MA_2

5

Slotted ALOHA (cont.)

1),,1(),0(

0)],,1(1)[,0(),0(),1(

1)],,0(1)[,1(

)(2),,(

,

inQnQ

inQnQnQnQ

inQnQ

nminiQ

P

ra

rara

ra

a

inn

),1(),0(),0(),1( nQnQnQnQP rarasucc

Page 6: 4_MA_2

6

Throughput of Slotted ALOHA

GeP 0

• The probability of no collision is given by

GeGPGS 0

• The throughput S is

368.01

max e

S

• The Maximum throughput of slotted ALOHA is

Page 7: 4_MA_2

7

ALOHA

1 2 3 3 2Time

Collision

Retransmission Retransmission

Node 1 Packet Waiting a random time

Node 2 Packet

Node 3 Packet

Page 8: 4_MA_2

8

Throughput of ALOHA

n

!n

(2G)nP

e 2G

• The probability that n packets arrive in two packets time is given by

where G is traffic load.

GeP 20

• The probability P(0) that a packet is successfully received without collision is calculated by letting n=0 in the above equation. We get

GeGPGS 20

• We can calculate throughput S with a traffic load G as follows:

184.02

1max

eS

• The Maximum throughput of ALOHA is

Page 9: 4_MA_2

9

Comparison of Aloha and S-Aloha

G86420

0.5

0.4

0.3

0.2

0.1

0

Slotted Aloha

Aloha

0.368

0.184

G

S

Page 10: 4_MA_2

10

CSMA: Carrier Sense Multiple Access

Page 11: 4_MA_2

11

Contention Protocols

CSMA (Carrier Sense Multiple Access) Improvement: Start transmission only if no

transmission is ongoing

CSMA/CD (CSMA with Collision Detection) Improvement: Stop ongoing transmission if a

collision is detected

CSMA/CA (CSMA with Collision Avoidance) Improvement: Wait a random time and try again

when carrier is quiet. If still quiet, then transmit

Page 12: 4_MA_2

12

Carrier Sense Multiple Access

In many multiaccess systems--e.g., LANs--ready station can determine if medium is idle before transmitting if medium is sensed as busy, ready station defers until

it becomes idle collisions are still possible if two (or more) ready

stations sense idle at same time

1 2 3

TimeCollision

4

Node 4 sense

Delay

5

Node 5 sense

Delay

Node 1 PacketNode 2 Packet

Node 3 Packet

Page 13: 4_MA_2

13

CSMA

Page 14: 4_MA_2

14

CSMA Slotted Aloha

The major difference between CSMA Slotted Aloha and ordinary slotted Aloha is that idle slots in CSMA have a duration β.

If a packet arrives at a node while a transmission is in progress, the packet is regarded as backlogged and begins transmission with probability qr after each subsequent idle slot.

Packets arriving during an idle slot are transmitted in the next slot as usual.

a.k.a. nonpersistent CSMA

Page 15: 4_MA_2

15

nonpersistent CSMA

Idle Period

Busy Period

Collision!!

Time

Page 16: 4_MA_2

16

CSMA Slotted Aloha Variations

persistent CSMA frames arriving during an idle slot β are transmitted at end of

the minislot arrivals during busy period are transmitted as soon as

medium is sensed as idle (after β) backlogged stations (holding collided frames) retransmit at

end of each idle minislot with probability qr

P-Persistent CSMA frames arriving during an idle minislot are transmitted at end

of the minislot arrivals during busy period are transmitted at end of each

idle minislot with probability p backlogged stations retransmit at end of each idle minislot

with probability qr < p

Page 17: 4_MA_2

17

Mathematical analysis of nonpersistent

Markov chain model (discrete time) state is number n of backlogged stations

each busy (success or collision) slot has unit length

each busy slot is followed by one (idle) minislot

each time step in the MC corresponds to a real time interval of either if no station transmits) or 1+ if at least one station transmits)

Page 18: 4_MA_2

18

CSMA Slotted Aloha Analysis

At a transition into state n (i.e., at the end of an idle slot), the prob. of no transmissions in the following slot is e-λβ(1-qr)n. The first term is the prob of no arrivals in the

previous idle slot The second term is the prob of no

transmissions by the backlogged nodes

Page 19: 4_MA_2

19

CSMA Slotted Aloha Analysis (cont.)

The expected time (T) between state transitions in the state n is β+(1-e-λβ(1-qr)n).

Clearly, β T ≦ ≦ β+1

Using Little’s Theorem, the expected number of arrivals between state transitions is:

E{arrival} = λ (β+1-e-λβ(1-qr)n)

Page 20: 4_MA_2

20

CSMA Slotted Aloha Analysis (cont.)

The expected number of departure between state transitions in state n is simply the probability of a successful transmission, that is given by:

The drift in state n is defined as the expected number of arrivals less the expected number of departures between state transitions.

nr

r

rnrn qe

q

nqqeD )1(

1)1(1

nr

r

r

rn

rn

r

succ

qeq

nq

qqneqe

PPP

)1(1

)1(!0

)()1(

!1

)(

]retx one arrival, no[]retx no arrival, one[

101

Page 21: 4_MA_2

21

CSMA Slotted Aloha Analysis (cont.)

For small qr, (1- qr)n-1 (1- q≒ r)n e≒ -qrn

Therefore,

where g(n) = λβ+ qrn is the expected number of attempted transmissions following a transition to state n

The drift is negative if

The numerator is the expected number of departures per state transition, and the denominator is the expected duration of a state transition period; thus the ratio can be interpreted as departure rate.

)()( )()1( ngngn engeD

)(

)(

1

)(ng

ng

e

eng

Page 22: 4_MA_2

22

Departure Rate (i.e., throughput)

g

g

e

ge

121

1

2g

λArrival rate

Departure rate:

Equilibrium

large backlog

Page 23: 4_MA_2

23

Throughput vs β

Using GNUPlot

4.4.2: skip

Page 24: 4_MA_2

24

CSMA unslotted Aloha

When a packet arrives, the transmission starts immediately if the channel is sensed to be idle.

If the channel is sensed to be busy, or if the transmission results in a collision, the packet is regarded as backlogged.

Each backlogged packet repeatedly attempts to retransmit at randomly selected times separated by independent, exponentially distributed random delays τ, with prob density xe-xτ

Page 25: 4_MA_2

25

CSMA unslotted Aloha (cont.)

We assume a propagation and detection delay of β, so that if one transmission starts at time t, another node will not detect that the channel is busy until t+β, thus causing the possibility of collisions.

Consider an idle period that starts with a backlog of n. The time until the first transmission starts is an exponentially distributed R.V. with rate G(n)=λ+nx

G(n) is the attempt rate in packets per unit time.

Page 26: 4_MA_2

26

CSMA unslotted Aloha (cont.)

A collision occurs if the next sensing is done within time β. Thus, the prob that this busy period is a collision is 1-e-βG(n)

The prob of a transmission following an idle period is e-βG(n)

The expected time from the beginning of one idle period until the next is 1/G(n) + (1+ β) The first term is the expected time until the first

transmission starts The second term is the time until the first

transmission ends and the channel is detected as being idle again.

Page 27: 4_MA_2

27

CSMA unslotted Aloha (cont.)

The departure rate when the backlog is n is given by:

For small β, the maximum value occurs when G(n)≒β-1/2, and the value is

The MAX value is slightly smaller than the MAX value of CSMA slotted Aloha. The reason is when CSMA is not being used, collisions are somewhat more likely fit a given attempt rate in an unslotted system than a slotted system.

)1()(/1

)(

nG

e nG

21

1

Page 28: 4_MA_2

28

CSMA unslotted Aloha (cont.)

However, in a slotted system, β would have to be larger than in an unslotted system to compensate for synchronization inaccuracies and worst-case propagation delay.

Thus, unslotted Aloha appears to be the natural choice for CSMA.

4.4.4: skip

Page 29: 4_MA_2

29

CSMA/CD: CSMA + Collision detection

Page 30: 4_MA_2

30

CSMA/CD In CSMA protocols

If two stations begin transmitting at the same time, each will transmit its complete packet, thus wasting the channel for an entire packet time

In CSMA/CD protocols The transmission is terminated immediately upon the

detection of a collision CD = Collision Detect

Page 31: 4_MA_2

31

CSMA/CD

Page 32: 4_MA_2

32

CSMA/CD (cont’d) Sense the channel

If idle, transmit immediately If busy, wait until the channel becomes idle

Collision detection Abort a transmission immediately if a collision is

detected Try again later after waiting a random amount of time

Page 33: 4_MA_2

33

CSMA/CD (cont’d) Carrier sense

reduces the number of collisions

Collision detection reduces the effect of collisions, making the

channel ready to use sooner

Page 34: 4_MA_2

34

Slotted CSMA/CD

We visualize S-CSMA/CD in terms of slots and minislots.

The minislots are of duration β, which denotes the time required for a signal to propagate from one end of the cable to the other and to be detected.

If the nodes are all synchronized into minislots of this duration, and if one node transmits in a minislot, all the other nodes will detect the transmission and not use subsequent minislots until the entire packet is completed.

Page 35: 4_MA_2

35

Slotted CSMA/CD (cont.)

If more than one node transmits in a minislot, each transmitting node will detect the condition by the end of the minislot and cease transmitting.

Thus, the minislots are used in a contention mode, and when a successful transmission occurs in a minislot, it effectively reserves the channel for the completion of the packet.

Page 36: 4_MA_2

36

Slotted CSMA/CD (cont.)

We assume each backlogged node transmits after each idle slot with prob qr

The node transmitting rate after an idle slot is Poisson with parameter g(n)=λβ+ qrn

Consider state transitions at the ends of idle slots: if no transmissions occur, the next idle slot ends after time β; if one transmission occurs, the next idle slot ends after 1+ β

Page 37: 4_MA_2

37

Slotted CSMA/CD (cont.)

Variable-length packets are allowed here, but the packet durations should be multiples of the idle slot durations.

For simplicity, we assume the expected packet duration is 1.

Finally, if a collision occurs, the next idle slot ends after 2β, i.e. nodes must hear an idle slot after the collision to know that it is safe to transmit.

Page 38: 4_MA_2

38

Slotted CSMA/CD (cont.)

The expected length of the interval between state transitions is: E{interval}=β+g(n)e-g(n)+β[1-(1+g(n)) e-g(n)] The second term is 1 times the success prob The third term is the additional β times the

collision prob

The prob of success is g(n)e-g(n)

The drift in state n is λE{interval} - Psucc

Page 39: 4_MA_2

39

Slotted CSMA/CD (cont.)

The departure rate in state n is

This quantity is maximized over g(n) at g(n)=0.77, and the resulting value is 1/(1+3.31β)

The constant (i.e. 3.31) is dependent on the detailed assumptions of the system. However, if β is very small, this constant is not very important.

Unslotted CSMA/CS makes more sense due to the difficulty of perfect synchronizing on short minislots.

)()(

)(

))(1(1)(

)(ngng

ng

engeng

eng

Page 40: 4_MA_2

40

Unslotted CSMA/CD

Suppose a node at one end starts to transmit, and then, almost β time units later, a node at the other end starts. The 2nd node ceases its transmission almost immediately upon hearing the 1st node, but nonetheless causes errors in the first packet and forces the 1st node to stop transmission another β time units later.

Node 1 starts

Node 2 starts Node 1 heardNode 2 stops

Node 2 heardNode 1 stops

TimePropagation

delay

Page 41: 4_MA_2

41

Unslotted CSMA/CD (cont.)

Nodes closer together on the cable detect collisions faster than those more spread apart.

As a result, the MAX throughput achievable with Ethernet depends on the arrangement of nodes on the cable and is very complex to calculate exactly.

Goal: to find bounds on all the relevant parameters from the end of one transmission to the end of the next in order to get a conservative bound on max throughput!

Page 42: 4_MA_2

42

Unslotted CSMA/CD (cont.) Assume that each node initiates transmissions according to an

independent Poisson process whenever it senses the channel idle, and the overall rate is G.

All nodes sense the beginning of an idle period at most β after the end of a transmission.

The expected time to the beginning of the next transmission is at most 1/G.

The next packet will collide with some later starting packet with prob at most 1-e-βg

The colliding packet will cease transmission after at most 2β

The packet will be successful with prob at least e-βg and will occupy 1 time unit.

Page 43: 4_MA_2

43

Unslotted CSMA/CD (cont.)

The departure rate S for a given G is the success prob divided by the expected time of a success or collision; so

The MAX occurs at

The corresponding MAX value is

This analysis is very conservative, but if β is very small, throughputs very close to 1 can be achieved.

GG

G

eeG

eS

)1(2/1

43.06

113

G

2.61

1

S

Page 44: 4_MA_2

44

Unslotted CSMA/CD (cont.)

The MAX stable throughput approaches 1 with decreasing β; whereas the approach is as a constant times β1/2 for CSMA.

The reason for the difference is that collisions are not very costly with CSMA/CD, and thus much higher attempt rates can be used.

CSMA/CD (and CSMA) becomes increasingly inefficient with increasing bus length (i.e. β), with increasing data rate (i.e. C), and with decreasing data packet size (i.e. L). ps:

L

C

Page 45: 4_MA_2

45

IEEE 802 LANs LAN: Local Area Network What is a local area network?

A LAN is a network that resides in a geographically restricted area

LANs usually span a building or a campus

Page 46: 4_MA_2

46

Characteristics of LANs

Short propagation delays

Small number of users

Single shared medium (usually)

Inexpensive

Page 47: 4_MA_2

47

Common LANs Bus-based LANs

Ethernet (*) Token Bus (*)

Ring-based LANs Token Ring (*)

Switch-based LANs Switched Ethernet ATM LANs

(*) IEEE 802 LANs

Page 48: 4_MA_2

48

OSI Layers and IEEE 802

802.2 Logical Link Control

802.3 802.4 802.5Medium Access Control

Data Link Layer

Physical Layer

Higher Layers

OSI layers IEEE 802 LAN standards

Higher Layers

CSMA/CD Token-passing Token-passing bus bus ring

Page 49: 4_MA_2

49

IEEE 802 Standards802.1: Introduction

802.2: Logical Link Control (LLC)

802.3: CSMA/CD (Ethernet)

802.4: Token Bus

802.5: Token Ring

802.6: DQDB

802.11: CSMA/CA (Wireless LAN)

Page 50: 4_MA_2

50

Summary

4.5.1, 4.5.3, 4.5.4, 4.5.5, 4.5.6, 4.6: skip