Message-ID: <4501AC5F.5080103@datatechlabs.com>
Date: Fri, 08 Sep 2006 17:46:07 +0000
From: alvis <alvis@datatechlabs.com>
User-Agent: Thunderbird 1.5.0.5 (X11/20060831)
MIME-Version: 1.0
To: Brian Wipf <brian@clickspace.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: Configuring System for Speed
References: <8FE8AE4A-43AD-4AE5-82B7-6C77BC4A0248@clickspace.com>
In-Reply-To: <8FE8AE4A-43AD-4AE5-82B7-6C77BC4A0248@clickspace.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Brian Wipf wrote:
> I am in the process of speccing out a new box for a highly utilized 
> (updates, inserts, selects) 40GB+ database. I'm trying to maximize 
> performance on a budget, and I would appreciate any feedback on any of 
> the following.
>
Perhaps this is off topic, but here is bit from my experience.  Using 
single server for both read (select) and write (insert, update, delete) 
operations is the way  to slow things down.  Consider to split query 
workload into OLTP and OLAP queries . Set up Slony replication and use 
slave server for read operations only . Buy 1 cheap box (slave) and 
another more expensive one for master. Keep in mind that DB schema 
optimization for particular query workload is essential . You just can 
not get good performance for selects if schema is optimized for inserts 
and vice versa.
>
> OS:
> The consensus in the list seems to be as long as you have the 2.6 
> Linux Kernel, it's really a matter of personal preference. However, 
> it's hard to have a preference when you're new to the Linux world, 
> like I am. Red Hat, Fedora Core, Slackware, Suse, Gentoo? I guess my 
> primary goal is speed, stability, and ease of use. Any advice here, no 
> matter how minimal, would be appreciated.
My answer is FreeBSD6.1.
>
> Thanks,
>
> Brian Wipf
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 6: explain analyze is your friend


-- 
Best Regards,
Alvis