xcp

From: Lillie Kittredge (kittredl@u.washington.edu)
Date: Tue Oct 19 2004 - 23:37:38 PDT

  • Next message: Tyler Robison: "Review for "Congestion Control for High Bandwidth-Delay Product Networks""

    Congestion Control Xtreem!

    This paper discusses the current state of, and proposes a new generation
    of, congestion control.

    The driving force for this paper is the increase in both high-bandwidth
    and high-delay network technologies. This represents a change in the
    environment since the earlier congestion control papers we've been
    reading. This again shows how congestion control has evolved with the
    available technologies; now a given packet traversing a network may
    encounter any number of different media, which may differ drastically in
    bandwidth and delay. One of the clever ideas of the authors for
    addressing this is a system of feedback information in packet headers
    which is updated by the routers along the packet's path; this ends up
    recording the information from the bottleneck, so the source (once the
    packet is returned to it) knows what it's really dealing with. This is a
    much more powerful meme than the source merely getting one of its packets
    dropped when the network is congested: now it knows not only that there is
    congestion, but exactly what it needs to do to avoid contributing to it.

    The main insight of the paper, which I rather like, is that fairness
    control should be separated from efficiency control. As such they create
    both a fairness and an efficiency manager for their protocol. The
    congestion manager uses the cool trick with the traffic feedback in the
    header; the fairness manager shuffles the available bandwidth among the
    flows.

    I like this paper and think it's got good ideas for how we can deal with
    the continued evolution of the Internet. I find it ironic that network
    technologies have not only gotten much wider - bandwidth, but have in some
    cases gotten slower (or at least with larger delays), like wifi. That
    seems an interesting and unexpected situation; it makes me wonder what
    other unexpected changes in network behavior are in our future.


  • Next message: Tyler Robison: "Review for "Congestion Control for High Bandwidth-Delay Product Networks""

    This archive was generated by hypermail 2.1.6 : Tue Oct 19 2004 - 23:37:39 PDT