From: David V. Winkler (dwinkler_at_windows.microsoft.com)
Date: Mon Mar 08 2004 - 17:51:17 PST
Review: Measurement, Modeling, and Analysis of a Peer-to-Peer File Sharing Workload
I attended the symposium presentation of this paper about a year and a half ago, and felt that I got more out of it than the paper itself.
The main point made by the paper is that the workload of a peer to peer system does not follow a Zipf curve, and the paper goes to great length explaining this. The reasons for this include the fetch-at-most-once nature of the network. 'the primary forces in Kazaa are the creation of new objects and the addition of new users.' I felt like more could be done to shoehorn this into a Zipf curve.
Good data is presented showing that Kazaa users are patient, much much more patient than web users.
The possibility of a cache of Kazaa users is discussed. The obvious solution of a proxy server at the edge of the university is dismissed for what seem like good reasons. The redirection answer seems like a good alternative.
This archive was generated by hypermail 2.1.6 : Mon Mar 08 2004 - 17:51:25 PST