X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id E8F5ED1B2BD for ; Wed, 21 Jul 2004 18:07:45 -0300 (ADT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 59743-01 for ; Wed, 21 Jul 2004 21:07:43 +0000 (GMT) Received: from mpls-qmqp-03.inet.qwest.net (mpls-qmqp-03.inet.qwest.net [63.231.195.114]) by svr1.postgresql.org (Postfix) with SMTP id C104ED1B220 for ; Wed, 21 Jul 2004 18:07:35 -0300 (ADT) Received: (qmail 10613 invoked by uid 0); 21 Jul 2004 20:36:54 -0000 Received: from mpls-pop-11.inet.qwest.net (63.231.195.11) by mpls-qmqp-03.inet.qwest.net with QMQP; 21 Jul 2004 20:36:54 -0000 Received: from 63-227-127-37.dnvr.qwest.net (HELO ?10.0.0.2?) (63.227.127.37) by mpls-pop-11.inet.qwest.net with SMTP; 21 Jul 2004 21:07:33 -0000 Date: Wed, 21 Jul 2004 15:08:30 -0600 Message-Id: <1090444110.709.38.camel@localhost.localdomain> From: "Scott Marlowe" To: "joe" Cc: "Postgresql Performance" Subject: Re: Beowulf Cluster & Postgresql? In-Reply-To: <1090442756.7202.38.camel@localhost> References: <1090442756.7202.38.camel@localhost> Content-Type: text/plain Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.1 tagged_above=0.0 required=5.0 tests=RCVD_IN_SORBS X-Spam-Level: X-Archive-Number: 200407/144 X-Sequence-Number: 7535 On Wed, 2004-07-21 at 14:45, joe wrote: > Hi all, > I was wondering if part or all of Postgres would be able to take > advantage of a beowulf cluster to increase performance? If not then why > not, and if so then how would/could it benefit from being on a cluster? > > Thanks for the enlightenment in advance. That type of clustering helps with large parallel processes that are loosely interrelated or none at all. In PostgreSQL, as in most databases, all actions that change the data in the database tend to be highly interrelated, so it becomes very expensive to pass all that locking information back and forth. The very thing a cluster would be good at, lots of reads, very few writies, is the antithesis of what postgresql is built to be good at, lots of writes as well as lots of reads. Basically, clustering tends to make the database faster at reads and slower at writes. While there are clustering solutions out there, Beowulf clustering is oriented towards highly parallel CPU intensive workloads, while PostgreSQL tends to be I/O intensive, and since all the data needs to be stored in one "master" place, adding nodes doesn't usually help with making writes faster.