Review: Binary Feedback for Congestion Avoidance

TCP is becoming inefficient and prone to instability as the
bandwidth-delay product is increasing. TCP congestion control mechanisms
are one source of this inefficiency. Different congestion control
mechanisms have been purposed. They focus on recovering from the
congestion but once congestion has developed a lot of effort is wasted to
restore the network. In order to avoid such a burden the author has
presented a scheme that tends to avoid congestion from developing in the
network. Author has also highlighted the analyses that were carried out to
decide between different design decisions for the scheme.

The congestion avoidance scheme presented in this paper is based on the
binary feedback and is applicable to connectionless networks and transport
protocol using window flow control. This purpose is to operate the network
at highly desirable rate instead of recovering when congestion has
occurred. In order to achieve this goal, the network provides explicit
feedback by setting a field in the network layer header to indicate the
congestion.

In order to carry out the design decision analyses and test the congestion
avoidance technique, a detailed simulation was used. In this simulation
computer network was modeled as multiple users generating packets in a
closed queuing network. Each hop is modeled by a single service time that
is different across different hops to have non-homogenous paths.

Using these simulations the author has shown that following results were
achieved

1. Congestion avoidance scheme (CAS) behaved well for random sized packets
and the dynamic behavior of window size was reasonable
2. CAS was able to scale when under transients behaviors
3. CAS adapts dynamically and adjust the window size to its correct value
in situation where the users start transmitting at arbitrary window sizes
or the network is overloaded.

Though the result are convincing but it would have been really helpful if
a performance comparison result between the network that have the
congestion avoidance scheme and those without congestion avoidance were
presented. Also the author has claimed that the scheme also maintains
fairness but has not provided any evidence to support the claim.
Received on Tue Sep 26 2006 - 01:02:22 EDT