X-Original-To: pgsql-performance@postgresql.org Received: from email01.aon.at (WARSL401PIP8.highway.telekom.at [195.3.96.97]) by postgresql.org (Postfix) with SMTP id 5668F474E4F for ; Wed, 19 Mar 2003 14:53:11 -0500 (EST) Received: (qmail 620460 invoked from network); 19 Mar 2003 19:53:10 -0000 Received: from m169p021.dipool.highway.telekom.at (HELO cantor) ([62.46.11.21]) (envelope-sender ) by qmail1rs.highway.telekom.at (qmail-ldap-1.03) with SMTP for ; 19 Mar 2003 19:53:10 -0000 From: Manfred Koizar To: "Kendrick C. Wilson" Cc: chriskl@familyhealth.com.au, pgsql-performance@postgresql.org Subject: Re: postgresql meltdown on PlanetMath.org Date: Wed, 19 Mar 2003 20:52:21 +0100 Message-ID: <00ih7vcuu5gf29oep2hi9sctrnui061ogh@4ax.com> References: In-Reply-To: X-Mailer: Forte Agent 1.8/32.548 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-Archive-Number: 200303/137 X-Sequence-Number: 1435 On Tue, 18 Mar 2003 09:19:44 -0600, "Kendrick C. Wilson" wrote: >If there are multiple values, the location of the first record is found in >the indexFile. > >Then dataFile is scanned until this != 'whatever'; Nice, but unfortunately not true for Postgres. When you do the first UPDATE after CLUSTER the new version of the changed row(s) are written to the end of the dataFile (heap relation in Postgres speech). So the *index* has to be scanned until this != 'whatever'. >Clustering is good for queries that return multiple [rows with the same search] values. Yes. With clustering you can expect that most of the tuples you want are near to each other and you find several of them in the same page. Servus Manfred