Re: Measuring Performance/Tuning - Mailing list pgsql-general

From Francisco J Reyes
Subject Re: Measuring Performance/Tuning
Date
Msg-id 20030211195441.A19529-100000@zoraida.natserv.net
Whole thread Raw
In response to Measuring Performance/Tuning  (Hunter Hillegas <lists@lastonepicked.com>)
List pgsql-general
On Tue, 11 Feb 2003, Hunter Hillegas wrote:

> When watching postmaster in top, I see it go from 4% to 30% of the CPU and
> back and forth, etc...

I don't think CPU utilization is a good measure of the need to change
parameters. You could have a badly configured setup which may not make the
CPU go high depending on the queries. You could have the same too.. a
poorly configured server which even with simple queries makes the CPU go
high.

> > Is there an easy way to determine if I need to adjust some settings,
> sort_mem, etc... to boost performance? Or trial and error?

I think trial and error is your ultimate solution.
What I do is that I have a table where I store the start/stop for most of
my routinarily runned queries. If I want to see if a setting change helps
I just make the change and run any of the daily queries. I have months
worth of sample times including changing machine, OS, etc.. I also may put
a note on the log such as "CPU/Memory upgrades" in the log so I can have
an idea of when I expect to see changes.


pgsql-general by date:

Previous
From: Stephane Schildknecht
Date:
Subject: Postgresql Partnership
Next
From: Stephan Szabo
Date:
Subject: Re: CREATE TABLE with a name derived from a string