Review of Internet Indirection Infrastructure

From: Ethan Katz-Bassett (ethan@cs.washington.edu)
Date: Sun Nov 14 2004 - 23:32:34 PST

  • Next message: Tyler Robison: "Review of "Internet Indirection Infrastructure""

    In the paper, the authors present their Internet Indirection Infrastructure,
    an overlay to provide rendezvous-based communication on the Internet. As we
    have seen in class, the Internet protocols were designed for unicast
    point-to-point communication between fixed endpoints. Because of this
    design focus, the protocols are not ideal for dealing with other situations,
    such as multicast. Solutions to problems such as multicast have required
    some amount of jerry rigging, with researchers either arguing for changes to
    the underlying protocols or giving schemes implemented in end hosts or by
    overlays. Generally, these solutions do not attempt to be general purpose.

     

    The Internet Indirection Infrastructure is a general overlay service (on top
    of IP) that can be used to solve many of these problems. Basically,
    receivers tell the service to send packets with specific IDs to its address,
    and senders give the service ID and data packet pairs. The service then
    sends (over IP) the data to the addresses of any hosts interested in that
    ID. The system also allows for ID stacks and for partial matching of IDs.
    They then describe how this simple system enables multicast, mobility of
    hosts, and other services. For instance, the system does not need to
    explicitly implement multicast trees; ID stacks can be used for this goal.
    I found the explanation of service composition to be particularly
    interesting; I had not thought about this problem before.

     

    The authors did a great job of addressing most of the concerns I had while
    reading the paper. I have numerous questions written in the margins, then
    scribbled out as they were answered in later questions: how the random IDs
    work, whether or how nodes pick nearby i3 servers, what happens when an i3
    server goes down, etc. The sampling scheme to pick a nearby server seemed
    somewhat clumsy, but, otherwise, I was impressed with how they cleanly
    addressed so many issues. Latency stretch occurred, but was limited.

     

    The Internet Indirection Infrastructure has a simple design that seems
    capable of offering clean solutions to many Internet problems. I am curious
    to see what other services it can enable and also what it is incapable of.

     


  • Next message: Tyler Robison: "Review of "Internet Indirection Infrastructure""

    This archive was generated by hypermail 2.1.6 : Sun Nov 14 2004 - 23:32:39 PST