1

An Adaptive Polling Algorithm for PCF Mode of 802.11 Wireless LANs

Xuanming DongPravin Varaiya

Anuj Puri

Speaker ： Mark Yang

2

Outline

Abstract

Introduction

Adaptive Polling Algorithm

Simulation

Conclusion

Reference

3

Abstract

Proposes an adaptive polling algorithm to improve the

wireless medium utilization.

The priority for each mobile station will be dynamically

updated using the Additive Increase/Multiplicative Decrease

algorithm.

The performance of wireless LANs are improved in terms of

the successful poll rate and the aggregate throughput.

4

Introduction – PCF mode

DCF (Distributed Coordination Function) – Contention ModePCF (Point Coordination Function) – Contention Free Mode

PCF most suitable for time-bounded applications. Polling frame

Stations are allowed to send data only when they receive the polling frames from the AP.

If the AP enables the PCF mode, the stations still can decide whether they want to be polled or not.

Round Robin polling algorithm Traffic is not evenly distributed over all mobile stations

→ Inefficient scheduling.

5

Introduction – Polling list

Only the stations in the polling list maintained by the point

coordinator are eligible to receive polls.

Station can ask to be appended to the polling list by sending the

Association Request frame.

If a station is in the polling list initially but it doesn’t want to be

polled late, it can send Reassociation Request frame to the point

coordinator.

6

Introduction – Superframe

7

Adaptive Polling Algorithm– Basic Notations

Definition: Successful poll Missed poll

Good polling algorithms: Keep the medium busy for payload data instead of

management and control frames Keep the medium shared by all stations fairly.

Adaptive polling algorithm: based on recent poll feedbacks from wireless stations to

dynamically update priority Improve the overall throughput and successful poll rate.

8

Adaptive Polling Algorithm– Dynamic priority Assignment (1)

A successful poll means more potential traffic from a station. Priority of a station is related to the amount of traffic from it. Additive Increase / Multiplicative Decrease (AIMD)

Highest Priority

Lowest Priority

9

Adaptive Polling Algorithm– Dynamic priority Assignment (2)

Mac-802_11.cc

10

Adaptive Polling Algorithm– Dynamic priority Assignment (3)

If a station has bursty traffic, it takes at most log2m consecutive polls for the station to reach the highest priority .

Log28 = 3

11

Adaptive Polling Algorithm– Poll scheduling Algorithm (1)

All stations with the same priority are placed in the same list.

12

Adaptive Polling Algorithm– Poll scheduling Algorithm (2)

In round i , stations with priorities from i to 1 will be polled.

Priority = 1 ….. i ….. mpoll slot = m …. m-i+1 ….. 1

13

Adaptive Polling Algorithm– Poll scheduling Algorithm (3)

1 1 1 1 1

2 2 2 2

3 3 3

4 4

5

14

Adaptive Polling Algorithm– Poll scheduling Algorithm (4)

Mac-802_11.cc

15

Adaptive Polling Algorithm– Debug & Verify

1

1 1 1 1 1

2 2 2 2

3 3 3

4 4

5 1, 2, 3, 4, 5, 6, 7, 8, 9, 10

1

16

Simulation – Setup (1)

Software: Linux 6.2 + NS 2.1b8a + PCF support patch

Geography : AP connected to a wired node via a 10 mbps Ethernet link. The wireless LAN consists of the AP and 10 wireless stations. All stations and the AP are close enough to receive signals from

each other. In addition, all stations don’t move around. Traffic :

10 CBR flows from 10 different stations to the wired node through the AP.

These CBR flows send packets at rates of 200 packets per second with packet size 1000 bytes.

These flows are started one by one with a 2 second interval. They last 20 seconds and terminate.

17

Simulation – Setup (2)

TCL

18

Simulation – Trace (1)

namtrace

newtrace

19

Simulation – Trace (2)

Round robin Adaptive

Node ID Time

Traffic

20

Simulation – Performance (1)

X. Dong & P. Varaiya & A. Puri Mark Yang

21

Simulation – Performance (2)

X. Dong & P. Varaiya & A. Puri Mark Yang

22

Conclusion

This paper present an efficient polling algorithm based on

dynamic priority assigned by the point coordinator.

The priority of each station is updated regularly based on

the recent poll feedback from the station.

Proposed polling algorithm improves the performance of

wireless LANs in terms of the successful poll rate and the

aggregate throughput of AP.

Future work focuses on the theoretical analysis of the

new polling algorithm. It will also be interesting to study

how the DCF mode affects PCF performance.

23

References