On Oct 31, 10:02am, Pat Wilson wrote:
> Subject: efficient network topologies for AFS
>
> We're in the middle of a network redesign, and I'm reconsidering the
> deployment of our AFS servers. I've always assumed that server/client
> traffic exceeded server/server traffic such that it makes more sense to
> distribute servers in a single cell topologically closer to clients (i.e.
> scatter the servers across subnets) than it would to cluster the servers on
> a single (separate?) subnet - is there any evidence that this is _not_ so,
> for the common case?
>
> Are there any "best practices" papers on network topolgy?
>
> Thanks
>
> Pat Wilson
> [EMAIL PROTECTED]
>-- End of excerpt from Pat Wilson
Server to server traffic will depend upon how much you replicate
volumes and how often you update and release them. If you are
releasing big volumes frequently, the peak server to server traffic
can be substantial.
However, you still don't want to concentrate them in one place if
you care about fault tolerance. We have three data centers between
our NY and NJ sites. We've distributed the AFS servers among them
so that even a major outage at one of the three does not take out
the AFS service.
Andy
--
Andy Sherman 101 Hudson St, Jersey City NJ, 34th flr
Manager, Systems Administration (201) 524-5460
Lehman Brothers Global Unix Support [EMAIL PROTECTED]
"The point of technology is to serve the business, not the other way around."
