(no subject)

Summary of the paper

This paper presents the measured capacity of the Ethernet, both in
theory and in measurement. The main point is to clear some
misunderstanding of previous research work and suggest some ways to
implement to avoid future problems.

At first, this paper points out that most of theory analysis rely on
simplifying assumptions, either for tractability or simply to avoid
drowning in the variety of results that would come from exploring the
entire parameter space. After that, paper gives the parameters
affecting performance to show how performance is measured which
includes average delay, throughput and etc. Through the analysis of
previous theoretical studies, finally comes to several categories for
pitfalls for the unwary: Not the same protocol, unrealistic
assumptions, measuring the wrong dependent variable, using the wrong
independent variable, operating in the wrong regime. It's the "Myths
and reality".

In the third part of the paper, paper presents a measurements for a
real Ethernet. It shows several experiments and the result through
analysis and the figures which show the realistic characteristics of
the performance of Ethernet.

At last, paper gives the conclusion: The Ethernet is capable of
performance for typical of good performance for typical applications,
even at high offered load. Now the bad performance in the real world
is caused by misuse and misunderstanding of Ethernet, because it is
sensitive to packet size and the time it takes to detect a collision.

Here is some implications for ethernet implementations given in the
paper:

Don't install long cables

Don't put too many hosts on one cable

Implement the protocol correctly

Use the largest possible packet size

Don't mix serious real-time and serious bulkdata applications

Points in favour or against

The paper is generally well written, with fine and clear
presentation. Moreover, it's well organized. At the beginning, author
gives the question to be solved, then the brief guide of previous
study. For his own analysis and research work, it's mainly focused on
two parts: theoretical analysis and the measurements of a real
Ethernet. The most thing attracts me is the analysis "The Ethernet
was designed to handle bursty traffic with a high peak-to average
ratio of bit rates, rather than a continuous high load". That tells
me if you want to analysis a computer network, you should know the
design purpose and the characteristics of the network, then you can
get the correct conclusion and know how to implement application on
it. However, author did not give a theoretical analysis considering
the real parameter of Ethernet, like realistic assumptions, right
dependent variable, right independent variable.
Received on Sun Sep 17 2006 - 17:40:56 EDT