Thanks for all the responses. This has led me to a partial answer.

John's suggested tests yielded surprisingly high numbers. One thing
led to another, and I finally found that turning off compression
increased most clients' performance by a factor of 10. I must have
done something wrong when I tested this before, because I was still
getting the same crap results then.

I have a couple Windows clients that are still getting terrifically
bad results, but at least now I know that my server is not the
bottleneck.

A couple interesting notes:
 - Keeping your catalog database on the same drive/array as your
archive does not result in any performance loss IF you are running a
single concurrent job with data and attribute spooling enabled. The
spooling sequence is: spool data/attributes, despool data, despool
attributes. While watching my database queries/second, I saw zero
during the spooling and data despooling steps, and 600-1200/sec during
the despool attributes step. So this is a viable configuration for
those of you without separate disks for these things (separate
filesystems is still a good idea).

 - For me, decreasing compression (Compression = GZIP1) yielded zero
performance benefit. I was only able to increase performance by
turning it off altogether.

And one further question: I can GZIP some files and not others in a
fileset by using multiple Options resources, right?

Thanks again for all the help.

-HKS

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users