[OpenAFS] cache performance

[Phil.Moore@morganstanley.com]

>>>>> "Nathan" == Nathan Neulinger <nneul@umr.edu> writes:

Warren> This is still more authoratative for our purposes as there number of
Warren> clients is in the order of thousands (to tens of thousands) but the number
Warren> of servers is in the low hundreds.  We also control the builds on the
Warren> servers much, much more tightly, and can ensure that the audits run and
Warren> complete successfully.

Nathan> Out of curiosity, would y'all be willing to share some details
Nathan> of your architecture?

Actually, I have, several times.  

At LISA 95, Decorum 97, and several other conference keynote
presentations I've given.  Let me see if I can't dredge up something
and post it for external consumption this week.

We're huge.  By many metrics, the largest AFS shop on the planet, but
not the traditional metrics that are associated with scale.  We
probably have less data, fewer clients, and fewer servers than many
big engineering sites, but we have realtime mission critical
dependencies on AFS for nearly *ALL* production applications on UNIX
(and against my better judgement, soon Windoze as well), and when AFS
stops working, our business grinds to a halt.

I'll have more to say on this when I finish my coffee....