Date: Mon, 12 Dec 2005 13:37:38 -0700
From: Michael Fuhr <mike@fuhr.org>
To: Carlos Benkendorf <carlosbenkendorf@yahoo.com.br>,
	pgsql-performance@postgresql.org
Subject: Re: How much expensive are row level statistics?
Message-ID: <20051212203738.GA61439@winnie.fuhr.org>
References: <20051211115336.91041.qmail@web35506.mail.mud.yahoo.com>
	<20051211194442.GA97174@winnie.fuhr.org>
	<20051212132342.GE19555@surnet.cl>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20051212132342.GE19555@surnet.cl>

On Mon, Dec 12, 2005 at 10:23:42AM -0300, Alvaro Herrera wrote:
> Michael Fuhr wrote:
> > The cost depends on your usage patterns.  I did tests with one of
> > my applications and saw no significant performance difference for
> > simple selects, but a series of insert/update/delete operations ran
> > about 30% slower when block- and row-level statistics were enabled
> > versus when the statistics collector was disabled.
> 
> This series of i/u/d operations ran with no sleep in between, right?
> I wouldn't expect a normal OLTP operation to be like this.  (If it is
> you have a serious shortage of hardware ...)

There's no sleeping but there is some client-side processing between
groups of i/u/d operations.  As I mentioned in another message, the
application reads a chunk of data from a stream, does a few i/u/d
operations to update the database, and repeats several thousand times.

The hardware is old but it's adequate for this application.  What
kind of overhead would you expect?

-- 
Michael Fuhr