Coda File System

Re: load balancing question

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Mon, 1 Nov 2004 23:33:05 -0500
On Mon, Nov 01, 2004 at 10:09:11AM -0500, Mike Branda wrote:
> and write back to 122 separate files.  They all need the same source
> information but read different segments of the file and output the
> corresponding data.  All that said, I'm looking at a cluster that can
> have replicated volumes and will load balance the requests for the

With Coda the file attributes are always fetched from all available
servers, clients are reponsible for detecting differences between the
replicas. The file contents are fetched from the server that is
estimated to be the fastest (on the assumption that our connection to
that server is the more reliable or that the server is least loaded).
It doesn't matter whether a client reads the whole file or only a little
part, we fetch the whole thing once and keep using the locally cached
copy until the original on the server is changed.

Jan
Received on 2004-11-01 23:33:54