XCP

From: Kate Everitt (kteveritt@yahoo.com)
Date: Tue Oct 19 2004 - 22:19:31 PDT

  • Next message: Alan L. Liu: "Review of the mice /sec paper"

    Congestion Control for High Bandwidth-Delay Product
    Networks
    Review: Katherine Everitt

    This paper presented XCP, a new eXplicit Control
    Protocol to replace TCP. The key aspects of XCP are
    that it is more stable and efficient when bandwidth
    and latency increases, and that it decouples fairness
    and efficency.

    The main advantage of XCP is that its parameters are
    constants, and thus independent of things such as the
    delay and capacity at the bottleneck. This makes it
    easier to calibrate and more likely to work in a
    variety of situations. It is faster at noticing
    congestion too, as the implicit “packet loss” signal
    takes a while to time out, and may also indicate a
    packet loss instead of congestion. XCP is faster to
    come back up to after a congestion signal, because it
    has more speed. Because the fairness and efficiency
    controllers are decoupled, it is easier to make a
    change to one of them without affecting the other, and
    you can have more specific controllers in any given
    situation.

    The analysis was very detailed, and it seems that XCP
    is a better and more robust protocol than TCP.
    However, to deploy it in practice would involve making
    changes to routers, no easy task. Also, this will be
    complicating the network and does not really fully
    address the problem of Denial of Service attacks.
    Overall this seemed a good solution to a problem that
    is not-quite-here-yet, but if we see it, we’ll know
    what to do.

                    
    _______________________________
    Do you Yahoo!?
    Declare Yourself - Register online to vote today!
    http://vote.yahoo.com


  • Next message: Alan L. Liu: "Review of the mice /sec paper"

    This archive was generated by hypermail 2.1.6 : Tue Oct 19 2004 - 22:19:37 PDT