Re: shared_buffers/effective_cache_size on 96GB server - Mailing list pgsql-performance

From Bruce Momjian
Subject Re: shared_buffers/effective_cache_size on 96GB server
Date
Msg-id 20121010175639.GH11892@momjian.us
Whole thread Raw
In response to Re: shared_buffers/effective_cache_size on 96GB server  (Claudio Freire <klaussfreire@gmail.com>)
List pgsql-performance
On Wed, Oct 10, 2012 at 02:05:20PM -0300, Claudio Freire wrote:
> On Wed, Oct 10, 2012 at 1:10 PM, Bruce Momjian <bruce@momjian.us> wrote:
> >> >shared_buffers = 10GB
> >>
> >> Generally going over 4GB for shared_buffers doesn't help.. some of
> >> the overhead of bgwriter and checkpoints is more or less linear in
> >> the size of shared_buffers ..
> >>
> >> >effective_cache_size = 90GB
> >>
> >> effective_cache_size should be ~75% of the RAM (if it's a dedicated server)
> >
> > Why guess?  Use 'free' to tell you the kernel cache size:
> >
> >         http://momjian.us/main/blogs/pgblog/2012.html#May_4_2012
>
> Why does nobody every mention that concurrent access has to be taken
> into account?
>
> Ie: if I expect concurrent access to 10 really big indices, I'll set
> effective_cache_size = free ram / 10

It is true that the estimate assumes a single session is using all the
cache, but I think that is based on the assumion is that there is a
major overlap between the cache needs of multiple sessions.

--
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + It's impossible for everything to be true. +


pgsql-performance by date:

Previous
From: Craig James
Date:
Subject: Re: Hyperthreading (was: Two identical systems, radically different performance)
Next
From: Claudio Freire
Date:
Subject: Re: Hyperthreading (was: Two identical systems, radically different performance)