Excerpts from transarc.external.info-afs: 16-Jun-93 Daemon access
[EMAIL PROTECTED] (799)
> BTW: Do you really want a central mail spool for AFS? Didn't CMU have
> problems with AMS because that all machines were looking at one common
> directory (which got updated frequently as well) which generates tons of
> callbacks?
There are probably lots of possible such directories, but only a subset
of the directories were looked at by each CM machine. If you're talking
about the central delivery queues, there can be any number of them and a
given CM needs only to get to one of them. Perhaps the problem was that
the queue to start with was picked randomly each time, whereas it might
have been better to bias the starting choice by source machine, so that
a given machine would tend to use only one of the queue directories.
Don't know.
Certainly any really central, R/W resource will accumulate lots of
callbacks, and this will be a bigger problem as your client base grows.
Craig
