I am running a trac server that feels slower than it ought to be. Are there benchmarks or load testers that can be quantitative about this, and that can measure where time goes?
I am running NetBSD 4 on xen (i386) apache 2.2 mod_wsgi pgsql 8.1 There are about 100 users with accounts of whom only a few are making requests at once. The machine has 1G of RAM and the disk is RAID-1 in the dom0 and I get 60 MB/s from it in the domU. I have configured apache to not make so many children, because I had lots of workers taking up RAM that weren't getting used. I did a 'vaccuum full analyze;' in pgsql, and probably should do this via cron daily; it seemed to help. Is updating pgsql to 8.3 advisable? Should I be looking at creating more indexes? I'm inclined to put some logging in trac around db queries, but it's hard to tell where the time is going. Things like getting a timeline take 5s (there are 13000 revisions in the svn, and probably 700 commits in the last 30 days), and a ticket query that returns 50 tickets takes 1s or so.
pgpkMFARnTIlY.pgp
Description: PGP signature
