From: Rosalia Tungaraza (rltungar@u.washington.edu)
Date: Sun Nov 07 2004 - 21:08:46 PST
This paper is an analysis of the utilization of content delivery system
across the UW campus. The authors examined three types of traffic:
Peer-to-peer (P2P), World Wide Web (WWW), and Content Delivery Networks
(CDNs) each with a significantly different architecture from the other two
and thus different traffic characteristics. In this study they found that
P2P traffic dominates the others in terms of HTTP bytes transferred, and
average size of documents and thus document/file transfer time. Moreover,
only a small number of clients and servers accounted for much of that type
of traffic.
One of the successes of this paper is that it sheds some light into the
fact that P2P content delivery may not be scalable in the long run (at
least for universities). As noted in the paper, that type of network
traffic tends to consume much more bandwidth within a given time period
compared to the other ones (e.g. WWW traffic) because it transports much
larger objects (usually video and audio files). The authors give a
concrete example of this by showing that over the period of their trace,
an average web client consumed 41.9MB of bandwidth while an average Kazaa
peer consumed 3.6 GB of bandwidth.
The only drawback of this paper is that it only has data taken from one
university. The authors' findings and recommendations would have been much
stronger had they obtained similar data from other universities.
One possible future work that I think should be done is to perform a
similar study at two or more other universities and compare the results.
Though I find this very unlikely, UW might be an exception to the rule.
Even if that is not the case, we would have a confirmation about the
current trend of content delivery system utilization. I am also very
interested in knowing how other non-academic institutions utilize their
content delivery systems.
This archive was generated by hypermail 2.1.6 : Sun Nov 07 2004 - 21:08:55 PST