From: Daniel Lowd (lowd@cs.washington.edu)
Date: Sun Nov 07 2004 - 20:21:51 PST
This paper looked at traffic patterns of the WWW, content delivery
networks (Akamai), and peer-to-peer networks (Kazaa and Gnutella), at UW.
The news that P2P consumes a great deal of bandwidth should not be a
surprise, but it was nice to see it described in detail, quantitatively.
What was more interesting was that a few files consumed so much bandwidth,
and the suggestion of cacheing P2P files.
While the idea is novel, I doubt it would survive the first round of RIAA
and MPAA lawsuits. If students are distributing 600mb files themselves,
then the university can maintain plausible deniability in the event of
(likely) copyright infringement. If the university's ISP, however, begins
cacheing these files, then it might be held liable as a party to this
infringement. Forwarding bytes is one thing; keeping complete copies of
the files looks more suspicious, even if broken into 32kb blocks.
In fact, these worries over lawsuits are probably what contribute to the
poor scalability of P2P in the first place. If no user feared lawsuits,
many more would probably be willing to share the files they downloaded,
acting as local caches automatically. A poorly organized network might
still result in wasted bandwidth (downloading files from unnecessarily
far away), but if there were enough copies, this should be reduced.
However, I suspect that P2P networks are currently suffering from a "free
rider" problem, exacerbated by legal threats: no one has been sued for
downloading files, only for sharing them. Furthermore, those who share
the most files are most likely to be sued. Fearless users remain, and
those few do most of the serving; the rest take advantage of them, while
protecting themselves or their bandwidth.
Relevant future work would be to see if these patterns are consistent at
other universities, and how they change over time.
-- Daniel
This archive was generated by hypermail 2.1.6 : Sun Nov 07 2004 - 20:21:52 PST