MailScanner load

[Jeremy Blonde]

I've been wanting to use Bayes for our mailscanner system, but I'm
wondering now if the overhead is worth it.  We have a new mailscanner
box running with Postfix (under Gentoo Linux) and it works pretty well,
we're filtering between 75-80% of our mail as spam.  The box however, is
hammered, it's running around 5-6 with spikes to 8-9.  I've just turned
off bayes and I'll see how the load is once it's processed all the back
logged messages.  We're averaging about 15,000-20,000 messages a day
(probably more once the school year starts).  Mailscanner is running on
an HP DL320, hyper-threaded proc, 1 GB of RAM, and raid 1 drives.

Is that normal for using bayes with mailscanner or do I need to tweak
some things?

(I'm already using tmpfs for a little bit of a speed up).

Jeremy Blonde
Instructional Technology - Server Support
Grant Joint Union School District