From: Jonas Lindberg (jonaslin@kth.se)
Date: Wed Nov 03 2004 - 08:39:55 PST
Review of S. Saroiu, K. P. Gummadi, R. J. Dunn, S. D. Gribble, and H. M.
Levy's "An Analysis of Internet Content Delivery Systems"
By: Jonas Lindberg
In this paper the authors analyze inbound and outbound Internet traffic in
the University of Washington WAN for four different content delivery
systems: HTTP web, Akamai, Gnutella and Kazaa. Some results are interesting
and some fairly surprising. One of more interesting observations is how the
peer-to-peer traffic does not scale very well. The analysis shows that a
small number of Kazaa peers are responsible for a large amount of the total
Kazaa traffic, which is also substantial part of the total Internet traffic.
Another interesting part of the paper is where the authors suggest and
experiment with a cache for peer-to-peer applications. The experiments show
that using a P2P cache can result in a significant decreased in bandwidth
usage. It is certainly an interesting idea, but I wonder if this would be
possible to implement for a university, since it is essentially supporting
activities that in some cases violate copyrights.
The authors ambition in this paper seem to be to examine the general use of
content delivery systems in the Internet, and, weather or not this be the
case, it would have been interesting if they had commented on if the
collected data is representative for the rest of the Internet.
I think the paper has some interesting points, and the results presented are
well explained and illustrated in a good way. However, the paper still feels
a little thin and I do not think I have learned as much from it as from the
other papers we have read in this course.
This archive was generated by hypermail 2.1.6 : Wed Nov 03 2004 - 17:40:29 PST