Review: Storage management and caching in PAST, a large-scale, persistent peer-to-peer storage utility.

From: David V. Winkler (dwinkler_at_windows.microsoft.com)
Date: Wed Mar 03 2004 - 16:43:06 PST

  • Next message: Slavik Krassovsky: "A. Rowstron and P. Druschel. Storage management and caching in PAST, a large-scale, persistent peer-to-peer storage utility."

    Review: Storage management and caching in PAST, a large-scale, persistent peer-to-peer storage utility.

    PAST is a overlay over Pastry that provides routing.

    PAST achieves robustness through duplication of files. The files are distrubted according to hash of the filename, the owner public key, and a salt. The node id is similarly generated by a hash of the node's public key, so "there is no correlation between the value of the nodeId and the node's geographic location, network connectivity, ownership, or jurisdiction".

    PAST places a file on several servers whose nodeId is closest in distance to the fileId. This is elegant.

    Caching of popular data in unused portions of the network, is additionally done to speed up queries.

    The pastry discussion is pretty confusing, but that's not what the paper is really about.

    Replica diversion and file diversion allow for nodes without sufficient storage to hand off the responsibility of storing the file or storing the replica to nearby nodes.

    Section 3.6, File encoding, was to me the most exiting portion of the paper. It is an idea of using error correcting codes to allow files to be broken up around the network. An almost implementation is discussed in the CFS paper.

    The experimental results are necessary, but not interesting.


  • Next message: Slavik Krassovsky: "A. Rowstron and P. Druschel. Storage management and caching in PAST, a large-scale, persistent peer-to-peer storage utility."

    This archive was generated by hypermail 2.1.6 : Wed Mar 03 2004 - 16:43:15 PST