On Sun, 19 May 2002 00:31:49 -0700
"Cindy" <ctmoore@uci.edu> wrote:
> If we search on an item that is not to be found in our entries,
> steph goes through the postgres entries faster than att, perhaps
> by about 10% (not a markedly faster rate, but distinctly faster).
> If we search on an item that is commonly found and display 100
> results at a time on each page, then what takes att 5 seconds to
> generate and display takes steph 30 seconds to do the same thing.
This sounds like you might be sorting the data (of course, without
the actual query or query plan, it's difficult to tell); there was
a performance issue with Sun's implementation of qsort() analyzed
on the list recently -- check the archives.
> Is there any indication that 7.2 is slower overall than 7.0.3?
Not that I'm aware of (the inverse should be true).
Can you confirm that the relevant performance settings in postgresql.conf
are set to reasonable values? (shared_buffers, wal_buffers, sort_mem,
fsync, etc.)
Can you post the EXPLAIN results for the 30-sec/5-sec query on 7.0.3,
the EXPLAIN ANALYZE results for 7.2, and the relevant bits of the schema?
Cheers,
Neil
--
Neil Conway <neilconway@rogers.com>
PGP Key ID: DB3C29FC