On Thu, 13 Nov 2008 09:11:05 -0500
Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Teodor Sigaev <teodor@sigaev.ru> writes:
> >> Yeah, I'm not convinced either. Still, Teodor's theory should
> >> be easily testable: set synchronize_seqscans to FALSE and see
> >> if the problem goes away.
>
> > Test suit to reproduce the problem:
>
> I don't doubt that you're describing a real effect, I'm just not
> sure yet that it's the same thing Ivan is seeing. He seems to be
> talking about more than 4x differences.
I just tested dropping the index and recreating it on a quite slower
box (Core Duo 2 notebook with 1G RAM) and absolutely no tuning on
postgres.conf but a "pure" lenny box and not an etch with backported
postgresql.
It seems a bit faster than the fastest time I've been able to obtain
on a 2x Xeon HT 3.2GHz, 4Gb RAM and SCSI RAID1.
It's far from being a scientific measure. I'll try to do more
experiments later to collect more data and see if it didn't happen
by chance.
--
Ivan Sergio Borgonovo
http://www.webthatworks.it