To: "Jim C. Nasby" <jnasby@pervasive.com>
Cc: Simon Riggs <simon@2ndquadrant.com>,  Tom Lane <tgl@sss.pgh.pa.us>,
	Qingqing Zhou <zhouqq@cs.toronto.edu>,  pgsql-hackers@postgresql.org
Subject: Re: Stats collector performance improvement
References: <200601021840.k02Ieed21704@candle.pha.pa.us>
	<11924.1136233224@sss.pgh.pa.us> <dpc4cg$lm6$1@news.hub.org>
	<12503.1136238525@sss.pgh.pa.us>
	<1136281253.5052.113.camel@localhost.localdomain>
	<20060103163556.GH82560@pervasive.com>
In-Reply-To: <20060103163556.GH82560@pervasive.com>
From: Greg Stark <gsstark@mit.edu>
Organization: The Emacs Conspiracy; member since 1992
Date: 03 Jan 2006 18:28:34 -0500
Message-ID: <878xtw6gd9.fsf@stark.xeocode.com>
Lines: 24
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.4
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii


"Jim C. Nasby" <jnasby@pervasive.com> writes:

> I would argue that minutes is too long, but of course this could be
> user-adjustable. I suspect that even waiting just a second could be a
> huge win, since this only matters if you're executing a lot of
> statements and you won't be doing that if those statements are taking
> more than a second or two to execute.

That's not necessarily true at all. You could just as easily have a
performance problem caused by a quick statement that is being executed many
times as a slow statement that is being executed few times.

That is, you could be executing dozens of queries that take seconds or minutes
once a second but none of those might be the problem. The problem might be the
query that's taking only 300ms that you're executing hundreds of of times a
minute.

Moreover, if you're not gathering stats for queries that are fast then how
will you know whether they're performing properly when you look at them when
they do show up?

-- 
greg