Date: Tue, 27 Mar 2012 14:58:22 -0500 (CDT)
From: Joshua Berkus <josh@agliodbs.com>
To: Peter van Hardenberg <pvh@pvh.ca>
Cc: postgres performance list <pgsql-performance@postgresql.org>
Message-ID: <1125348837.158829.1332878302273.JavaMail.root@mail-1.01.com>
In-Reply-To: 
 <CAAcg=kUN57mRjKXcVHO_7on+924CyrcPep9wMHbuCSwWxRZ-8A@mail.gmail.com>
Subject: Re: Determining working set size
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Peter,

Check out pg_fincore.  Still kind of risky on a production server, but does an excellent job of measuring page access on Linux.

----- Original Message -----
> Baron Swartz's recent post [1] on working set size got me to
> thinking.
> I'm well aware of how I can tell when my database's working set
> exceeds available memory (cache hit rate plummets, performance
> collapses), but it's less clear how I could predict when this might
> occur.
> 
> Baron's proposed method for defining working set size is interesting.
> Quoth:
> 
> > Quantifying the working set size is probably best done as a
> > percentile over time.
> > We can define the 1-hour 99th percentile working set size as the
> > portion of the data
> > to which 99% of the accesses are made over an hour, for example.
> 
> I'm not sure whether it would be possible to calculate that today in
> Postgres. Does anyone have any advice?
> 
> Best regards,
> Peter
> 
> [1]:
> http://www.fusionio.com/blog/will-fusionio-make-my-database-faster-percona-guest-blog/
> 
> --
> Peter van Hardenberg
> San Francisco, California
> "Everything was beautiful, and nothing hurt." -- Kurt Vonnegut
> 
> --
> Sent via pgsql-performance mailing list
> (pgsql-performance@postgresql.org)
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-performance
>