Date: Sun, 29 May 2005 11:33:12 -0500
From: "Jim C. Nasby" <decibel@decibel.org>
To: Christopher Kings-Lynne <chriskl@familyhealth.com.au>
Cc: mark durrant <markd89@yahoo.com>,
	pgsql-performance@postgresql.org
Subject: Re: Select performance vs. mssql
Message-ID: <20050529163312.GK44623@decibel.org>
References: <20050524153637.64525.qmail@web53101.mail.yahoo.com>
	<4293D500.8010908@familyhealth.com.au>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <4293D500.8010908@familyhealth.com.au>
User-Agent: Mutt/1.5.8i

On Wed, May 25, 2005 at 09:29:36AM +0800, Christopher Kings-Lynne wrote:
> >--MSSQL's ability to hit the index only and not having
> >to go to the table itself results in a _big_
> >performance/efficiency gain. If someone who's in
> >development wants to pass this along, it would be a
> >nice addition to PostgreSQL sometime in the future.
> >I'd suspect that as well as making one query faster,
> >it would make everything else faster/more scalable as
> >the server load is so much less.
> 
> This is well-known and many databases do it.  However, due to MVCC 
> considerations in PostgreSQL, it's not feasible for us to implement it...

Wasn't there a plan to store some visibility info in indexes? IIRC the
idea was that a bit would be set in the index tuple indicating that all
transactions that wouldn't be able to see that index value were
complete, meaning that there was no reason to hit the heap for that
tuple.

I looked on the TODO but didn't see this, maybe it fell through the
cracks?
-- 
Jim C. Nasby, Database Consultant               decibel@decibel.org 
Give your computer some brain candy! www.distributed.net Team #1828

Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"