SpaceBallOne wrote:
> Hello everyone,
>
> First time poster to the mailing list here.
>
> We have been running pgsql for about a year now at a pretty basic
> level (I guess) as a backend for custom
> web (intranet) application software. Our database so far is a "huge"
> (note sarcasm) 10 Mb containing of about 6 or so principle tables.
>
> Our 'test' screen we've been using loads a 600kb HTML document which
> is basically a summary of our client's orders. It took originally 11.5
> seconds to load in internet explorer (all 10.99 seconds were pretty
> much taken up by postgres processes on a freebsd server).
>
> I then re-wrote the page to use a single select query to call all the
> information needed by PHP to draw the screen. That managed to shave it
> down to 3.5 seconds... but this so far is as fast as I can get the
> page to load. Have tried vacuuming and creating indexes but to no
> avail. (increasing shared mem buffers yet to be done)
>
> Now heres the funny bit ...
>
> Every time I tested an idea to speed it up, I got exactly the same
> loading time on a Athlon 1800+, 256Mb RAM, 20Gb PATA computer as
> compared to a Dual Opteron 246, 1Gb RAM, 70Gb WD Raptor SATA server.
> Now, why a dual opteron machine can't perform any faster than a lowly
> 1800+ athlon in numerous tests is completely beyond me .. increased
> memory and RAID 0 disc configurations so far have not resulted in any
> significant performance gain in the opteron server.
>
> Do these facts sound right? If postgres is meant to be a 200Gb
> industrial strength database, should it really be taking this long
> pulling 600kb worth of info from a 10Mb database? And why no
> performance difference between two vastly different hardware spec'd
> computers??? Am I missing some vital postgres.conf setting??
>
> Any advice welcome.
>
> Thanks,
> Dave
> space_ball_one@hotmail.com <mailto:space_ball_one@hotmail.com>
>
Could you give us a bit more info.
What you are trying to do. EXPLAIN ANALYZE would be great.
In my experience first problem with the first db app is no indexes used
in joining.
--
-- Andrei Reinus