Date: Wed, 28 Jun 2006 10:08:15 -0500
From: "Jim C. Nasby" <jnasby@pervasive.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Andrew Sagulin <andrews42@yandex.ru>, pgsql-performance@postgresql.org
Subject: Re: Large index scan perfomance and indexCorrelation (PG 8.1.4 Win32)
Message-ID: <20060628150815.GT44573@pervasive.com>
References: <463024718.20060627161408@yandex.ru>
	<1151445858.2479.170.camel@localhost.localdomain>
	<624168784.20060628133355@yandex.ru>
	<15072.1151505444@sss.pgh.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <15072.1151505444@sss.pgh.pa.us>
User-Agent: Mutt/1.5.11

On Wed, Jun 28, 2006 at 10:37:24AM -0400, Tom Lane wrote:
> with a plain indexscan.  What you need to do is compare the two
> functions and figure out what part of the cost models are out of line
> with reality.  I tend to agree with the upthread comment that the
> nonlinear interpolation between min_IO_cost and max_IO_cost is suspect

If you're going to make such a comparison (which is badly needed, imho),
http://stats.distributed.net/~decibel/ might be of use to you. It shows
that the nonlinear interpolation between the correlated and uncorrelated
index scans is way off base, at least for this example.

BTW, you'll have a hard time convincing people to increase the cost
estimates of index scans, because experience has shown that they're
already too high (all the discussions about dropping random_page_cost,
for example).
-- 
Jim C. Nasby, Sr. Engineering Consultant      jnasby@pervasive.com
Pervasive Software      http://pervasive.com    work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf       cell: 512-569-9461