"Wide Area Cooperative Storage with CFS" Review

From: Tarik Nesh-Nash (tarikn_at_microsoft.com)
Date: Wed Mar 03 2004 - 15:40:46 PST

  • Next message: David Coleman: "CFS review"

    This paper presents the Cooperative file system. A P2P readonly storage
    utility that offers an attractive solution to enhance the fault
    tolerance and load balance and other challenges that other P2P
    implementation suffer from.

    CFS is structured as a collection of servers that provide block level
    storage. The client interface is a readonly UNIX like interface. The
    publisher, however, can update the content by inserting blocks using a
    hach identifiers and signing it. The system validates the authenticity
    of the publisher by validating its public key. At the lowest level, the
    chord layer locates blocks across the network. Every node is defined
    using a nodeid. The network traffic is minimized by selecting the
    successor node as the next one in a circular space. The node uses two
    data structures to perform lookup: The successor table and the finger
    table. The lookup runs in a big O of LogN. A special consideration is
    taken when a new node is added or deleted. The DHash layer stored and
    retrieves blocks, handles distribution, replication and caching. It
    uses the chord layer to locate the blocks.

    I feel that the implementation is more complicated than it should be.
    Moreover, this system is limited since it is read only. Its main
    advantage over the PAST is the ability to distribute big files
    efficiently. I m not familiar with the topic but I believe that a good
    practical approach will be to target all these old unused computers.

     

     

     


  • Next message: David Coleman: "CFS review"

    This archive was generated by hypermail 2.1.6 : Wed Mar 03 2004 - 15:42:37 PST