From: weibel@oclc.org
Date: Mon, 3 Oct 1994 09:26:46 -0400
Message-Id: <199410031326.JAA02646@ws02-00.rsch.oclc.org>
To: uri@bunyip.com
Subject: Re: No "TOP" of the docuverse [Was: URC usage scenarios ]
> I suggest that any scenario involving a "root" or "top" server is
> doomed to fail.
> If you rely on a "root" server, you create a
> single point of failure, which conflicts badly with the goal of
> high availability.
A "root" or "top" server can be a single *logical* server without being
constrained to a single physical machine, and without being constrained
by the single point of failure phenomenon. Technology to support such
services is well understood and deployed widely.
OCLC does this now and supports transaction rates upwards of 100/second
at an availability rate of greater than 99.9%, and this from a single
site. Scaling such a system to multiple sites and far greater transaction
capacity is well within current production practice.
stu