PostgreSQL not fully utilizing system resources?

From: "Gregory Stewart" <gstewart512(at)gmail(dot)com>
To: pgsql-performance(at)postgresql(dot)org
Subject: PostgreSQL not fully utilizing system resources?
Date: 2007-06-01 19:40:19
Message-ID: d9731dbe0706011240o11214bc4s92d0785ac35f66a@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Hello List,

We've been running PostgreSQL as our web application database for
almost a year and it has noticeably slowed down over last few months.

Our current setup and pgsql configuration looks like this:

8.1.2 on Ubuntu 4 on Opteron Dual Core with 2 GBytes RAM. This is a
dedicated DB server.

We currently have about 3.5 million rows in 91 tables. Besides the
requests coming from the web server, we have batch processes running
every 15 minutes from another internal machine that do a lot of
UPDATE, DELETE and INSERT queries on thousands of rows.

Many of the SELECT queries coming from the web server contain large
JOINS and aggregate calculations.

We are running a financial application which is very data intensive
and calculates a lot on the SQL side.

Anyways, watching the system processes we realized that PostgreSQL is
only using about 300 Mbytes for itself. Also, both cores are usually
maxed out to 100% usage.

Are we expecting too much from our server?

Our non-default configuration settings are:

max_connections = 100
shared_buffers = 17500
work_mem = 2048
maintenance_work_mem = 40000
max_fsm_pages = 35000
autovacuum = on

What can I do to make best use of my db server? Is our configuration
flawed? Or are we already at a point where we need consider clustering
/ load balancing?

Any ideas and suggestions are welcome.

Regards,
Gregory Stewart

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Luke Lonergan 2007-06-01 19:54:44 Re: Autodetect of software RAID1+0 fails
Previous Message Dimitri 2007-06-01 18:51:00 Re: Autodetect of software RAID1+0 fails