Received: from localhost (maia-1.hub.org [200.46.204.191]) by postgresql.org (Postfix) with ESMTP id 758699FA3AC for ; Wed, 10 Jan 2007 06:08:07 -0400 (AST) Received: from postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.204.191]) (amavisd-new, port 10024) with ESMTP id 56534-01 for ; Wed, 10 Jan 2007 06:07:58 -0400 (AST) X-Greylist: domain auto-whitelisted by SQLgrey-1.7.4 Received: from mx01.bfk.de (mx01.bfk.de [193.227.124.2]) by postgresql.org (Postfix) with ESMTP id 6551A9FA31E for ; Wed, 10 Jan 2007 05:34:43 -0400 (AST) Received: from [10.119.110.2] (helo=mx00.int.bfk.de) by mx01.bfk.de with esmtp id 1H4ZqE-0002M8-Lh; Wed, 10 Jan 2007 10:33:54 +0100 Received: from fweimer by bfk.de with local id 1H4ZqD-0007k8-2Y; Wed, 10 Jan 2007 10:33:53 +0100 To: "Jeremy Haile" Cc: pgsql-performance@postgresql.org Subject: Re: High inserts, bulk deletes - autovacuum vs scheduled vacuum References: <1168363601.5177.1168380907@webmail.messagingengine.com> <82d55okrim.fsf@mid.bfk.de> <1168373668.28926.1168412995@webmail.messagingengine.com> From: Florian Weimer Date: Wed, 10 Jan 2007 10:33:53 +0100 In-Reply-To: <1168373668.28926.1168412995@webmail.messagingengine.com> (Jeremy Haile's message of "Tue, 09 Jan 2007 15:14:28 -0500") Message-ID: <82vejfi5tq.fsf@mid.bfk.de> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable X-Virus-Scanned: Maia Mailguard 1.0.1 X-Archive-Number: 200701/129 X-Sequence-Number: 22730 * Jeremy Haile: > Good advice on the partitioning idea. I may have to restructure some of > my queries, since some of them query across the whole range - but it may > be a much more performant solution. How is the performance when > querying across a set of partitioned tables vs. querying on a single > table with all rows? Locality of access decreases, of course, and depending on your data size, you hit something like to 2 or 4 additional disk seeks per partition for index-based accesses. Sequential scans are not impacted. > Does my current approach of disabling autovacuum and manually vacuuming > once-an-hour sound like a good idea, or would I likely have better > results by auto-vacuuming and turning row-level stats back on? Sorry, I haven't got much experience with autovacuum, since most of other databases are INSERT-only (or get VACUUMed automatically after major updates). --=20 Florian Weimer BFK edv-consulting GmbH http://www.bfk.de/ Kriegsstra=DFe 100 tel: +49-721-96201-1 D-76133 Karlsruhe fax: +49-721-96201-99