Distributed Weighted Fair Queuing in 802'11 Wireless LAN

1
Distributed Weighted Fair Queuing in 802.11
Wireless LAN

• Albert Banchs
• Xavier Perez
• ICC 2002

2
Outline

• Introduction
• Distributed Weighted Fair Queuing (DWFQ)
• Simulations
• Conclusion

3
Introduction

• Weighted Fair Queuing algorithms on a wired link
  are the efficient way to providing desired
  bandwidth to a flow in proportion to its weight.
• The challenge in Wireless LAN
• In wired links, competing in centralized queue
  link gateway.
• Wireless hosts are distributed, and they compete
  with each other.
• The author proposed a fully distributed weighted
  fair queuing (DWFQ).
• Extending DCF of 802.11 MAC protocol

4
DWFQ - Concept

• In DCF of 802.11 MAC, bandwidth received by a
  node depends on its CW.
• The smaller the CW, the higher the throughput,
  and vice versa.
• EX

• DWFQ determines the CW values to control desired
  bandwidth distribution.

5
DWFQ

• Assumption one flow per node
• ri bandwidth of flow i (estimated)
• li pkt length ti inter-arrival time (?t
  between 2 pkt), K constant
• Wi weight of flow i (assigned, and static)
• Wi ? 1 (weight of original 802.11 1 ?
  compatible)
• S , put Li into pkt header
• Each node receives others L and compares with
  its L.
• To determine the value of its CW.

6
DWFQ
1 is smallest (original)
7
DWFQ - Overload

• Overload
• As many stations have small CWs, each one of them
  should receive large throughput.
• ? large packet in flight ? more collision times
• s

increase CW
8
DWFQ - Overload

• t is a constant.
• c is too low high weight cant get desired
  bandwidth.
• c is too large of collisions is very high.
• c 5 (by Simulation Results)

9
DWFQ Multiple Flows per Node

• S
• Between flows at a node, using a weighted fair
  queuing (WFQ) scheduler to choose the next pkt to
  transmit.

10
Simulations - Utility

• 2Mbps Wireless LAN
• 10 Stations
• 2 stations weight 2 8 stations weight 1
• UDP traffic per stations with packet size 1000
  bytes

11
Simulations Function of Weight

• HP in graph LP weight 1
• Experienced Weight throughput of HP /
  throughput of LP

12
Simulations Number of Stations

• HP weight 2 LP weight 1
• Experienced Weight throughput of HP /
  throughput of LP

13
Simulations impact of c

• Cant be dropped more than 8 RTS tries 0 lt c lt 8
• Experienced Weight throughput of HP /
  throughput of LP

choose c 5
14
Simulations TCP sources

• 2HP one TCP, one UDP 8LP all UDP

15
Conclusion

• This paper proposed a distributed WFQ in wireless
  LAN.
• It focuses on fairness between node and node.
• Each node maintains L to control the value of
  CW.
• DWFQ can be compatible with the 802.11 standard.