Review: Fair Qing Algo

Congestion is one of the problems that the currently fast growing
networks are facing. Different solutions have been purposed in order to
control congestion that could be categorized into two categories based
on their implementation at either source or gateway. In this paper the
author argues that queuing algorithms when implemented effectively
affect the collective behavior of flow control algorithms and thus form
an important component of congestion control mechanism. Based on this
argument author has purposed a Fair Queuing (FQ) algorithm.

FQ algorithm purposed in this paper is based on the observation that the
queuing algorithm should allocate the bandwidth and buffer space fairly
and in order to achieve this they must be combined with intelligent flow
control algorithms at source. The FQ algorithm presented in this paper
is a slight modification to Nagle's purposed fair queuing algorithm in
that it fetches packets from different queues considering the packet
size and the promptness requested instead of simple round robin. In
order to demonstrate the FQ algorithm author has used analysis and
simulation. Different combinations of flow control mechanism and queuing
algorithms were simulated at packet level using a network simulator
built on Nest network simulation tool. Using these simulations author
has measured throughput, average roundtrip time, packet retransmissions
and packets dropped and has presented his result.

The results show that FQ behaves better than FCFS queuing algorithm in
the scenarios of multi-hop networks consisting of client like FTP
servers (continuous large chunk of data) and Telnet servers (small data
packet at distinct intervals) both in assigning fair bandwidth and
improved roundtrip time. The FQ algorithm also protects from problems of
vulnerabilities to ill-behaved sources and the results have shown that
the ill-behaved clients suffer in terms of bandwidth and roundtrip time.
On the other hand the sources using less than their fair share of
bandwidth have better results in terms of low delays.

An FQ algorithm seems to have promising results but unless it has been
tested on a real network nothing can be predicted about its future. Also
the author has not explicitly the support that is required from the
gateways in order to implement it. These issues need to be addressed
before the FQ algorithms can become a standard queuing algorithm.
Received on Thu Sep 28 2006 - 10:27:38 EDT