Message-ID: <4E931846.80509@2ndQuadrant.com>
Date: Mon, 10 Oct 2011 12:07:34 -0400
From: Greg Smith <greg@2ndQuadrant.com>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
	rv:1.9.1.16) Gecko/20111004 Icedove/3.0.11
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: Adding more memory = hugh cpu load
References: <4E92F294.9010405@gmail.com> <4E92FB87.1030703@peak6.com>
In-Reply-To: <4E92FB87.1030703@peak6.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 10/10/2011 10:04 AM, Shaun Thomas wrote:
> The problem with having lots of memory is... that you have lots of 
> memory. The operating system likes to cache, and this includes writes. 
> Normally this isn't a problem, but with 48GB of RAM, the defaults (for 
> CentOS 5.5 in particular) are to use up to 40% of that to cache writes.

I make the same sort of tuning changes Shaun suggested on every CentOS 5 
system I come across.  That said, you should turn on log_checkpoints in 
your postgresql.conf and see whether the "sync=" numbers are high.  That 
will help prove or disprove that the slowdown you're seeing is from too 
much write caching.  You also may be able to improve that by adjusting 
checkpoint_segments/checkpoint_timeout, or *decreasing* shared_buffers.  
More about this at 
http://www.westnet.com/~gsmith/content/postgresql/chkp-bgw-83.htm

There are some other possibilities, too, like that memory addition can 
actually causing average memory speed to drop as Kevin mentioned.  I 
always benchmark with stream-scaling:  
https://github.com/gregs1104/stream-scaling before and after a RAM size 
change, to see whether things are still as fast or not.  It's hard to do 
that in the position you're in now though.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us