(no subject)

Summary of the paper

This paper discusses gateway queueing algorithms and their role in
controlling congestion in datagram networks. It also compares this
algorithm to other congestion control schemes.

The main point of this paper indicates that queueing algorithms,
which control the order in which packets are sent and the usage of
the gateway's buffer space, do not affect congestion directly, but it
does determine the way in which packets from different sources
interact with each other which, in turn, affects the collective
behavior of flow control algorithms.It can protect the well-performed
the source and punish the ill-behaved source through which the
networks could be fair. The paper makes queueing algorithms a crucial
component in effective congestion control.

In the introduction, paper critics the previous fair queueing
algorithm in the gateway. They are not adequate and some lacks
implementation and testing to support the conclusion. Authors use the
modification of the algorithms and the aim is to find a queueing
algorithm that functions well in current computing environments.

In the second part, paper describes the new fair queueing algorithm.
Firstly, authors discuss the design requirement for an effective
queueing algorithm. The requirement is that the queueing algorithm
allocates bandwidth and buffer space fairly. Then, paper discusses
the definition of "fair". Finally, allocation on the basis of source-
destination pairs, or conversations, seems the best tradeoff between
security and efficiency and will be used in the new algorithm. Based
on the analysis above, authors give the definition of the algorithm.
They use a packet-by-packet transmission algorithm which is very
simple defined. Later on, authors analyzed some situation, such as
promptness, and the situation that queue is full and a new packet
arrives. Finally, authors try to characterize the promptness
allocation for an arbitrary arrival stream of packets using modeling
method, and present a delay-throughput curve given by the fair
queueing algorithm.

In the third part, authors discuss the flow control algorithm, and in
section 4, they present simulation data comparing several
combinations of flow control and queueing algorithm. The measurement
and analysis suggests that fair queueing algorithms make self-
optimizing source behavior result in fair, protective,
nonmanipulable, and stable networks; in fact, they may be the only
reasonable queueing algorithms to do so.

Points in favour or against

The paper is generally well written, with fine and clear
presentation. Obviously, the tests for this fair queueing algorithm
are limited, and are in no way conclusive. Moreover, as mentioned in
the paper, is the fair algorithm easy to deploy? You should consider
the cost of deployment. We can not violate the standard. It's true
that for individuals, they could alter the control algorithms on
their own machines, but is it necessary to do that?If only a small
part of Internet users deploy the algorithm, will it be also
effective? The fair algorithm this paper proposes works on the
gateway, so is it realistic to do that? I think there are tons of
questions that authors did not solve.
Received on Wed Sep 27 2006 - 17:22:29 EDT