ANNOUNCE: Passive Proximate Mirror Selection - Squid Patches

Folks,

As some of you may know, this spring AppliedTheory Communications
conducted an experiment to test the validity of using BGP AS Path
Lengths as a basic heuristic of 'network distance' when trying to
determine which geographically distributed mirror is optimal.

In April the results of that study were made publicly available.

The results provided a stronger than anticipated indication that this
is an effective strategy to facilitate proximate cluster selection
within mirrored resource environments.

The website for all materials related to this project is at
         http://proximate.appliedtheory.com

As a followup to the experiment we would like to make freely available
under a redistributable and modifiable software license, an
infrastructure for wider testing of this strategy.

Now available at the proximity web site:

  * A demonstration application to try out a few IP addresses and see
    how far they are perceived from the hosting web server network
    (located in Syracuse, New York). Install the Path Length
    Resolution Server yourself (it's free too!) and try it out from
    your own network.

  * Source code for a Path Length Resolution Server (requires a read
    only BGP feed, can translate requests for a set of IP addresses
    into a set of proximity metrics)

  * Patches for Squid 2.1 that enable squid to lookup a hostname in
    the global mirror directory service for a set of server IPs, and
    resolve those via the Path Length Resolution Server so web clients
    automatically take advantage of proximity resolution. If a host
    is not in the gmds, the server is selected according to normal
    dns.

  * An explanation of the global mirrored directory service (which is
    a DNS subtree for the registration of mirrored services) and an
    offer to provide DNS for your mirrors for you, or to delegate them
    to name servers that match the name servers for them in
    traditional root DNS space.

Still available at the web site

  * The original feasibility study, in a variety of document formats.

  * Open definitions of the protocol clients use to request proximity
    resolution services.

Questions, comments, and feedback eagerly appreciated!

Regards,
Patrick
Received on Wed Jun 02 1999 - 08:20:08 MDT