Received: from localhost (mx1.hub.org [200.46.208.251]) by postgresql.org (Postfix) with ESMTP id 384E49FB336 for ; Mon, 23 Oct 2006 19:08:52 -0300 (ADT) Received: from postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.208.251]) (amavisd-new, port 10024) with ESMTP id 39727-01 for ; Mon, 23 Oct 2006 19:08:37 -0300 (ADT) X-Greylist: from auto-whitelisted by SQLgrey- Received: from noel.decibel.org (noel.decibel.org [67.100.216.10]) by postgresql.org (Postfix) with ESMTP id A78959FB2CC for ; Mon, 23 Oct 2006 19:08:36 -0300 (ADT) Received: by noel.decibel.org (Postfix, from userid 1001) id 1FF4156427; Mon, 23 Oct 2006 17:08:35 -0500 (CDT) Received: (hashcash-sendmail, from uid 1001); Mon, 23 Oct 2006 17:08:33 -0500 Date: Mon, 23 Oct 2006 17:08:33 -0500 From: "Jim C. Nasby" To: John Philips Cc: pgsql-performance@postgresql.org Subject: Re: Optimizing disk throughput on quad Opteron Message-ID: <20061023220833.GX26892@nasby.net> References: <20061021154305.48104.qmail@web57806.mail.re3.yahoo.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20061021154305.48104.qmail@web57806.mail.re3.yahoo.com> X-Operating-System: FreeBSD 6.1-RELEASE-p10 amd64 X-Distributed: Join the Effort! http://www.distributed.net User-Agent: Mutt/1.5.13 (2006-08-11) X-Hashcash: 1:20:061023:johnphilips42@yahoo.com::EcunlOSh4d5/NNVT:00000000000000 0000000000000000000000002zX3 X-Hashcash: 1:20:061023:pgsql-performance@postgresql.org::qYNW5j58Re2FCTF8:00000 0000000000000000000000001UDk X-Virus-Scanned: Maia Mailguard 1.0.1 X-Spam-Status: No, hits=1.966 tagged_above=0 required=5 tests=AWL, SPF_HELO_SOFTFAIL, SPF_SOFTFAIL X-Spam-Level: * X-Archive-Number: 200610/492 X-Sequence-Number: 21621 On Sat, Oct 21, 2006 at 08:43:05AM -0700, John Philips wrote: > I heard some say that the transaction log should be on > it's own array, others say it doesn't hurt to have it > on the same array as the OS. Is it really worthwhile > to put it on it's own array? It all depends on the controller and how much non-WAL workload there is. Theoretically, with a good enough controller, you can leave WAL on the same partition as your data. With a complex setup like you're looking at, you really will want to do some testing to see what makes the most sense. I can also point you at a company that does modeling of stuff like this; they could actually give you some idea of how well that setup would perform before you buy the hardware. BTW, any test results you can provide back to the community would be most appreciated! -- Jim Nasby jim@nasby.net EnterpriseDB http://enterprisedb.com 512.569.9461 (cell)