From: Bruce Momjian <pgman@candle.pha.pa.us>
Message-Id: <200606161624.k5GGO3N03466@candle.pha.pa.us>
Subject: Re: Test request for Stats collector performance improvement
In-Reply-To: <4492D882.6030109@kaltenbrunner.cc>
To: Stefan Kaltenbrunner <stefan@kaltenbrunner.cc>
Date: Fri, 16 Jun 2006 12:24:03 -0400 (EDT)
CC: Qingqing Zhou <zhouqq@cs.toronto.edu>, pgsql-hackers@postgresql.org
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=US-ASCII

Stefan Kaltenbrunner wrote:
> Bruce Momjian wrote:
> > OK, based on reports I have seen, generally stats_query_string adds 50%
> > to the total runtime of a "SELECT 1" query, and the patch reduces the
> > overhead to 25%.
> 
> that is actually not true for both of the platforms(a slow OpenBSD
> 3.9/x86 and a very fast Linux/x86_64) I tested on. Both of them show
> virtually no improvement with the patch and even worst it causes
> considerable (negative) variance on at least the Linux box.

I see the results I suggested on OpenBSD that you reported.

> OpenBSD 3.9-current/x86:
> 
> without stats:
>     0m6.79s real     0m1.56s user     0m1.12s system
> 
> -HEAD + stats:
>     0m10.44s real     0m2.26s user     0m1.22s system
> 
> -HEAD + stats + patch:
>     0m10.68s real     0m2.16s user     0m1.36s system

and I got similar results reported from a Debian:

	Linux 2.6.16 on a single processor HT 2.8Ghz Pentium compiled
	with gcc 4.0.4.

	> > real        0m3.306s
	> > real        0m4.905s
	> > real        0m4.448s

I am unclear on the cuase for the widely varying results you saw in
Debian.

-- 
  Bruce Momjian   http://candle.pha.pa.us
  EnterpriseDB    http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +