CMPE 257: Wireless and Mobile Networking SET 3f:

Winter 2004 UCSC CMPE252B

1

CMPE 257: Wireless and Mobile Networking

SET 3f:

Medium Access Control Protocols

Spring 2005 CMPE257 UCSC 2

MAC Protocol Topics Fair medium access and fair

scheduling (queueing) MACAW Topology independent fair queuing


Unequal opportunity to access the channel

Severe throughput degradation Causes

Binary exponential backoff (BEB) Location dependent contention

The Fairness Problem

Node A

Node B

Node C Node D

Example: Flow CD will capture the channel.


Prior Work on Fairness Max-min fairness

Reduce the ratio between max throughput and min throughput of flows

Backoff or dynamic adjustment of channel access.

Fair queuing (FQ) approach Adapt wireline FQ disciplines to ad hoc networks. Flow contention graph as a useful abstraction Flows are tagged as either leading or lagging,

and then backoff window is adjusted accordingly.


MACAW ([BDSZ94]) One of the earliest work on throughput

and fairness enhancement Note: Packet sensing (not carrier sensing)

is used! Per-node and per-stream fairness

Maintain backoff windows for different streams (also appears in IEEE 802.11e)

Backoff copy and MILD backoff Multiplicative increase, linear decrease to

address the fairness problem


Topology Independent Fair Queuing Design Goals [LL00, LL05]

QoS Support for Advanced Applications in Ad Hoc Wireless Networks

Fair Allocation of Channel BandwidthMaximal Channel Utilization under Fairness

Constraint

– Communication Intensive Applications

– Limited-bandwidth Wireless Channel

Distributed Packet Scheduling Design


Outline Issues in Ad Hoc Wireless Fair

Queueing Fair Queueing Model for Packet

Scheduling An Idealized Centralized Algorithm A Distributed Implementation Performance Evaluation Conclusion and Future Work


Network Model A Single Shared Physical Channel Collision – Receiver in Transmission

Range of More than One Transmitting Node

Flow – Stream of Packets from Source to Destination <Sender, Receiver, Flow_ID>

CSMA/CA MAC Framework


Design Issues Location Dependent Contention &

Channel Reuse

Spatial Reuse

F1

F2

F3

X

Spatial Collision


Design Issues

F1

F2

F3

F4

No Spatial ContentionNo Spatial Reuse


Flow Contending Graph

F1

F2

F4

F3

F1F4

F2

F3

Flow GraphNode Graph


Design Issues (cont’d)

To Maximize Channel Utilization

F1

F2

F4F3 F5

F1

F2

F4

Node Graph

F3

Flow Graph

F5

• Inherent Conflict between Achieving Fairness & Maximizing Channel Utilization

F3F5

• Schedule F3 & F5 always (MIS)

F3 F5

F1

F2

F4

• Starve F1, F2 & F4

F1

F2

F4

Schedule the Maximal Number of Flows in Flow Graph


Design Issues (cont’d)

Distributed Nature of Packet Scheduling in Ad Hoc Wireless Networks Unlike Wireline or Packet Cellular Networks.

NO Single Logical Entity for Scheduling NO Direct Access to All Contending Flow

Info Provide QoS at Finest Time Scale

(Packet Level)


Solution Space Maximize Channel Utilization Always:

Schedule Largest Number of Non-conflicting Flows

Starvation of Certain Flows Ensure Fairness:

Maximize Spatial Channel Reuse Subject to Fairness Constraint


Enable Spatial Channel Reuse

F1 F2

F4 F3 Wireline FQ:0 1 2 3 4 5 6 7

F1.1 F2.2F2.1 F3.1 F4.1 F1.2 F3.2 F4.2

F1.1 F2.2F2.1 F3.1 F4.1 F1.2 F3.2 F4.2FQ Tagging:F1

F3


Look Ahead Window

F1 F2

F4 F3 Wireline FQ:0 1 2 3 4 5 6 7

F1.1 F2.2F2.1 F3.1 F4.1 F1.2 F3.2 F4.2

F1.1 F2.2F2.1 F3.1 F4.1 F1.2 F3.2 F4.2FQ Tagging:

FQ withLookahead:

F2.1F4.1

F1.2F3.2

F2.2F4.2

Lookahead window = 4 bits

F1.1F3.1

F1.1 F3.1


Minimum Graph Coloring Maximizing Spatial Reuse within a

Look Ahead Window is a Minimum Coloring Problem

F1

F2

F3

F5F0

F4

Flow Contending Graph

F1

F2

F3

F5F0

F4


Dynamic Graph Coloring Look Ahead Window Moves Forward Balance Two Design Goals:

Transmit Current Window of Bits ASAP Move Window Ahead AFAP


An Adaptive Algorithm Two Key Components:

A Basic Scheduling Loop Adaptive Dynamic Coloring


Basic Scheduling Loop

WFQ to Assign Flow Tags V(t): Smallest Finish Tag A Lookahead Window of Bits:

[V(t),V(t)+ ] Partition Flows into Disjoint Sets Schedule the Set with Least Tagged

Flow Move Window Forward to Next Least

Tagged Flow


Adaptive Dynamic Coloring As Lookahead Window Moves from

[V(t-1),V(t-1)+ ] to [V(t),V(t)+ ] Retain Disjoint Sets of Unserved Packets

in [V(t-1),V(t-1)+ ] Merge Newly Joined Packets, Create

New Set if No Merge Possible Retain All Disjoint Sets until t+1


Properties of Central Algorithms

Number of Disjoint Sets Non-increase Adaptively Reduce Total Number of

Disjoint Sets Move the Window Forward

Fairness Guarantee Spatial Channel Reuse Throughput & Delay Bounds


Distributed Implementation Approximate Central Algorithm: A Back-

off Based Approach Within CSMA/CA MAC Framework Approximate WFQ with Modified WRR Backoff-based Implementation of

Largest-degree First (LF) Coloring Algorithm

Backoff-based Implementation of Adaptive Algorithms


An Example

F0

F1

F2

F3

F4

F5

• Partition Flows into Disjoint Sets(5)

(2)

(2)

(2)

(3)

(4)

Adaptive Dynamic Coloring


An Example Largest Degree First:

Set Backoff = – Flow_Degree

F1

F2

F3

F4

F5F0(5)

(2)

(2)

(2)

(3)

(4)

Time

F2

F3

F4

F5

F0

F1

3

2

1

4

4

4


An Example

F1

F2

F3

F4

F5F0

Time

F2

F3

F4

F5

F0

F1

3

2

1

4

4

4

Wireline FQ:0 1 2 3 4 5 6 7

F1.1 F0.1F4.1 F3.1 F2.1 F5.1 F1.2 F4.2FQ Tagging:

F1.1F5.1

F1.1 F5.1


An Example

F1

F2

F3

F4

F5F0

Time

F2

F3

F4

F5

F0

F1

3

2

1

4

4

4

Wireline FQ:0 1 2 3 4 5 6 7

F0.1F4.1 F3.1 F2.1 F1.2 F4.2FQ Tagging:

F1.1F5.1

F4.1F0.1

F4.1 F0.1


An Example

F1

F2

F3

F4

F5F0

Time

F2

F3

F4

F5

F0

F1

3

2

1

4

4

4

Wireline FQ:0 1 2 3 4 5 6 7

F3.1 F2.1 F1.2 F4.2FQ Tagging:

F1.1F5.1

F4.1F0.1

F2.1F3.1

F3.1 F2.1

……

……


Other Issues Detailed MAC Layer Design

CSMA/CA Paradigm Global Flow Information (i.e. Flow

Weights) Propagation Conflict-free Multicast Tree


Simulation Example

F0

F2

F1

F3

F4

F5

F6

F7

F11

F15

F14

F13

F9

F8

F12

F10

F16

17 flows 15 source nodes Simulation slots:

100,000 Poisson & MMPP

Traffic


Simulation Example Normalized Throughput

0 2 4 6 8 10 12 14 160.8

0.85

0.9

0.95

1

1.05

Flow ID

Thr

ough

put

Normalized Throughput


Average Delay

0 2 4 6 8 10 12 14 160

0.5

1

1.5

2

2.5

3

3.5

4

Flow ID

De

lay

Average Delay

Average Delay


Convergence of Adaptive Coloring Numbers of Disjoint Sets

0 20 40 60 80 100

2

4

6

8

10

12

14

16

18

Time

Num

ber o

f Set

s

Number of Sets in WindowTotal Number of Sets


Related Work Distributed Fair Queuing

Adapt Fair Queuing Algorithm to Wireless LAN Distributed Fair Scheduling Backoff Based on Virtual Time No Explicit Consideration of Spatial Channel Reuse “Distributed Fair Scheduling in a Wireless LAN,” by

N. Vaidya, P. Bahl & S. Gupta, MOBICOM 2000 Multihop Packet Scheduling

Focus on Resolving Conflict between Fairness & Maximal Channel Utilization

Per-flow Service is not Fair “A New Model for Packet Scheduling in Multihop

Wireless Networks” by H. Luo & S. Lu, MOBICOM 2000


References [BDSZ94] Bharghavan et al., MACAW:

Media Access Protocol for Wireless LANs, in ACM SIGCOMM 1994.

[LL00] H. Luo and S. Lu, A Topology-Independent Fair Queueing Model in Ad Hoc Wireless Networks, in IEEE ICNP 2000.

[LL05] H. Luo and S. Lu, A Topology-Independent Wireless Fair Queueing Model in Ad Hoc Networks 2005, in IEEE JSAC 2005. (Extended version of [LL00]).


Acknowledgments

Parts of the presentation are adapted from the following sources: Prof. Luo’s ICNP 2000 presentation

Documents

CMPE 257: Wireless and Mobile Networking SET 3f: