Re: Monitoring number of backends

From: John R Pierce <pierce(at)hogranch(dot)com>
To: pgsql-general(at)postgresql(dot)org
Subject: Re: Monitoring number of backends
Date: 2013-10-22 19:02:56
Message-ID: 5266CBE0.1000703@hogranch.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

On 10/22/2013 10:59 AM, Stephen Frost wrote:
> PG is really*much* happier if you have only one backend per CPU in your
> system. The way to get there is by using a connection pooler like
> pg_bouncer and configuring it based on how many CPUs you have.

Actually, I've found peak throughputs on a decent multicore server with
lots of ram, and lots of disk IO parallelism (eg, big raid10) is aruond
2X the socket*hyperthread*core count... so for instance, on a modern 2
socket E5-2665 kind of server, thats 2 x 8 cores with 2 threads per
core, thats 16 total cores, 32 total hardware threads, so about 64
database connections would be peak, given a decent raid10 of lots of
SAS2 10k/15k disks

--
john r pierce 37N 122W
somewhere on the middle of the left coast

In response to

Responses

Browse pgsql-general by date

  From Date Subject
Next Message Moshe Jacobson 2013-10-22 19:15:41 Re: Bug? Function with side effects not evaluated in CTE
Previous Message John R Pierce 2013-10-22 18:59:42 Re: how to get the connected session pointer ( Archive * AH)