Re: Looking for tips

From: Richard Huxton <dev(at)archonet(dot)com>
To: Oliver Crosby <ryusei(at)gmail(dot)com>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: Looking for tips
Date: 2005-07-19 18:10:20
Message-ID: 42DD420C.2010306@archonet.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Oliver Crosby wrote:
> Hi,
> I'm running Postgres 7.4.6 on a dedicated server with about 1.5gigs of ram.
> Running scripts locally, it takes about 1.5x longer than mysql, and the load
> on the server is only about 21%.

What scripts? What do they do?
Oh, and 7.4.8 is the latest release - worth upgrading for the fixes.

> I upped the sort_mem to 8192 (kB), and shared_buffers and
> effective_cache_size to 65536 (512MB), but neither the timing nor the server
> load have changed at all.

Well, effective_cache_size is the amount of RAM being used by the OS to
cache your files, so take a look at top/free and set it based on that
(pick a steady load).

What sort_mem should be will obviously depend how much sorting you do.

Drop shared_buffers down to about 10000 - 20000 (at a guess)

You may find the following useful
http://www.varlena.com/varlena/GeneralBits/Tidbits/index.php
Read the Performance Tuning article, there is an updated one for version
8 at:
http://www.powerpostgresql.com/PerfList

> FYI, I'm going to be working on data sets in the
> order of GB.

Fair enough.

> I think I've gone about as far as I can with google.. can anybody give me
> some advice on how to improve the raw performance before I start looking at
> code changes?

Identify what the problem is first of all. Some things to consider:
- Are there particular queries giving you trouble?
- Is your load mostly reads or mostly writes?
- Do you have one user or 100?
- Are you block-loading data efficiently where necessary?
- Have you indexed both sides of your foreign-keys where sensible?
- Are your disks being used effectively?
- Are your statistics accurate/up to date?

Bear in mind that MySQL will probably be quicker for simple queries for
one user and always will be. If you have multiple users running a mix of
multi-table joins and updates then PG will have a chance to stretch its
legs.

--
Richard Huxton
Archonet Ltd

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Oliver Crosby 2005-07-19 18:21:22 Re: Looking for tips
Previous Message Tom Lane 2005-07-19 18:09:56 Re: Impact of checkpoint_segments under continual load conditions