Reviw: Congestion Avoidance and Control

Computer network are growing at a rapid pace but with this growth the
congestion in the network is also increasing and sometimes because of
this congestion the routers have to drop these packets. In order to
access this situation the author has carried out an investigation and
has implemented some new algorithms to enhance the TCP performance.

It all started in Oct 86 when the data throughput from LBL to UC Berkley
dropped from 32Kbps to 40bps and according to author this was because of
misbehaving of 4.3BSD TCP implementation. Based on these new algorithms
have been implemented in 4BSD TCP and these algorithms are round-trip
time variance estimation, exponential retransmit timer backoff, slow
start, aggressive receiver acknowledgement policy, dynamic window sizing
and fast retransmit. These algorithms have their respective tasks e.g.
slow start is being used either to start or restart after packet loss,
round-trip time variance estimation is being used to predict the round
trip time to avoid unnecessary retransmissions, exponential retransmit
timer backoff to stabilize the system in cases of retransmissions,
dynamic window sizing and fast retransmit to avoid congestion.

This is a well written paper that has addressed some of the issues that
resulted in congestion collapse and the results of the experiments
showed that the TCP with new algorithms implemented is better at dealing
with congestion on the internet.
Received on Tue Sep 26 2006 - 01:38:06 EDT