Message-Id: <4FABA1EE0200002500047AE0@gw.wicourts.gov>
Date: Thu, 10 May 2012 11:09:34 -0500
From: "Kevin Grittner" <Kevin.Grittner@wicourts.gov>
To: <pgsql-admin@postgresql.org>,<gnanam@zoniac.com>
Subject: Re: Advice/guideline on increasing shared_buffers and
	kernel parameters
References: <00da01cd2ce7$35d17780$a1746680$@com>
	<00e101cd2cfc$b6140b00$223c2100$@com>
	<4FA8E99B0200002500047A16@gw.wicourts.gov>
	<001f01cd2dd0$4187f8d0$c497ea70$@com>
	<4FAA3F740200002500047A73@gw.wicourts.gov>
	<005401cd2e7c$19fb1480$4df13d80$@com>
In-Reply-To: <005401cd2e7c$19fb1480$4df13d80$@com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

"Gnanakumar" <gnanam@zoniac.com> wrote:
 
> I've already thought of converting this into a document and keep
> it handy so that I may want to refer back whenever I need.
 
I've put up a first cut at such a document as a Wiki page:
 
http://wiki.postgresql.org/wiki/Number_Of_Database_Connections
 
Everyone should feel free to improve upon it.  I'll probably add a
"thought experiment" I've used a few times which seems to help some
people understand the issue.
 
>> A formula which has held up pretty well across a lot of
>> benchmarks for years is that for optimal throughput the number of
>> active connections should be somewhere near 
>> ((core_count * 2) + effective_spindle_count).
> 
> Our entire Production application stack is setup in Amazon EC2
> cloud environment, that includes database server also.  So, in
> that case, how do I find out "effective_spindle_count"?  I know
> "core_count" can be determined from Amazon EC2 instance type.  Per
> Amazon EC2, EBS volumes are reportedly a shared resource.
 
I think you need to experiment with different pools sizes.  Please
post results and/or update the Wiki page.
 
-Kevin