X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (av.hub.org [200.46.204.144]) by postgresql.org (Postfix) with ESMTP id 9529F9DC826 for ; Fri, 10 Feb 2006 18:42:26 -0400 (AST) Received: from postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 27143-02 for ; Fri, 10 Feb 2006 18:42:28 -0400 (AST) X-Greylist: from auto-whitelisted by SQLgrey- Received: from exchange.g2switchworks.com (mail.g2switchworks.com [63.87.162.25]) by postgresql.org (Postfix) with ESMTP id 2532E9DC800 for ; Fri, 10 Feb 2006 18:42:23 -0400 (AST) Received: from 10.10.1.37 ([10.10.1.37]) by exchange.g2switchworks.com ([10.10.1.2]) with Microsoft Exchange Server HTTP-DAV ; Fri, 10 Feb 2006 22:42:25 +0000 Received: from state.g2switchworks.com by mail.g2switchworks.com; 10 Feb 2006 16:42:25 -0600 Subject: Re: Large Database Design Help From: Scott Marlowe To: Ragnar Cc: pgsql-performance@postgresql.org In-Reply-To: <1139611194.17929.105.camel@localhost.localdomain> References: <43EB92DA.2080607@orangekids.org> <43EC69EA.3010709@logix-tt.com> <1139611194.17929.105.camel@localhost.localdomain> Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1139611345.22740.151.camel@state.g2switchworks.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.6 (1.4.6-2) Date: Fri, 10 Feb 2006 16:42:25 -0600 X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, score=0.157 required=5 tests=[AWL=0.156, UNPARSEABLE_RELAY=0.001] X-Spam-Score: 0.157 X-Spam-Level: X-Archive-Number: 200602/120 X-Sequence-Number: 17106 On Fri, 2006-02-10 at 16:39, Ragnar wrote: > On Fri, 2006-02-10 at 11:24 +0100, Markus Schaber wrote: > > > For lots non-read-only database workloads, RAID5 is a performance > > killer. Raid 1/0 might be better, or having two mirrors of two disks > > each, the first mirror holding system, swap, and the PostgreSQL WAL > > files, the second one holding the data. > > I was under the impression that it is preferable to keep the WAL on > its own spindles with no other activity there, to take full advantage > of the sequential nature of the WAL writes. > > That would mean one mirror for the WAL, and one for the rest. > This, of course, may sometimes be too much wasted disk space, as the WAL > typically will not use a whole disk, so you might partition this mirror > into a small ext2 filesystem for WAL, and use the rest for files seldom > accessed, such as backups. Well, on most database servers, the actual access to the OS and swap drives should drop to about zero over time, so this is a workable solution if you've only got enough drives / drive slots for two mirrors.