[e2e] Measurements of P2P traffic
David P. Reed
dpreed at reed.com
Wed Jul 31 07:21:11 PDT 2002
Many proposed P2P applications are moving to UDP. It makes a lot of sense.
Or "simulating UDP" on an overlay network. The "stream-wise" congestion
control mechanisms of TCP seem very unlikely to work on P2P real-time
collaborative or computer-computer apps (they work for real-time
collaborative apps based on a central server because of the central server).
It would be intellectually weak to extrapolate from Napster or Gnutella
architectures as if they are examplars of the future. Even weaker to
optimize for them. They are very special cases, with huge files dominating.
At 07:39 PM 7/30/2002 -0700, Yuchung Cheng wrote:
>I would appreciate if someone can point me any references or papers on
>measurements of current Peer-to-Peer traffic patterns in the Internet. For
>example, the percentage and behaviours of P2P routing, lookup/query
>traffic in the Internet. I am interested in knowing how the current and
>proposed End-to-End TCP/IP implementation/design will impact P2P network
>and the Internet.
>Execuse me for posting a not so e2e related article here. I can't find
>any P2P mailing list, is there any?
>CSE dept, UC San Deigo
More information about the end2end-interest