From: shearerje_at_comcast.net
Date: Thu Mar 04 2004 - 08:44:38 PST
CFS is distributed read-only file system using a non-hierarchical (peer-to-peer) confederation of member nodes. It uses a three layer hierarchy consisting of (1) the Chord distributed hash table for managing its logical topology, selection of which nodes will store a particular file, and routing algorithms, (2) DHash for managing the storage of files broken into data blocks, and (3) FS for presenting a file-centric interface to the client applications.
CFS is similar to PAST in its use of public/private keys for data naming and authentication, and in its enforcement of quotas for publishers. It’s odd that PAST identifies itself as write-once while CFS identifies itself as read-only. As near as I can tell they offer the same level of writablity.
CFS differs from PAST in that it breaks arbitrary size files into fixed size blocks that may or may not be stored on the same node. This simplifies load balancing but complicates file retrieval. CFS also imposes a storage time interval on content, making it more of a giant cache than a permenant data archive.
Much of the paper focuses on the characteristics of the Chord layer, which was covered very well by Krishna Gummadi in his comparison of DHT implementations (e.g., PASTRY’s logical address tree structure vs. Chord’s logical ring structure).
This archive was generated by hypermail 2.1.6 : Thu Mar 04 2004 - 08:44:44 PST