X-Original-To: pgsql-general-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [64.117.224.130]) by svr1.postgresql.org (Postfix) with ESMTP id BB418D1B4F1 for ; Mon, 15 Sep 2003 14:41:58 +0000 (GMT) Received: from svr1.postgresql.org ([64.117.224.193]) by localhost (neptune.hub.org [64.117.224.130]) (amavisd-new, port 10024) with ESMTP id 83770-04 for ; Mon, 15 Sep 2003 11:41:30 -0300 (ADT) Received: from internetstaff.com (internetstaff.com [63.214.174.240]) by svr1.postgresql.org (Postfix) with ESMTP id 251A9D1B4EE for ; Mon, 15 Sep 2003 11:41:26 -0300 (ADT) Received: from spank (ip68-110-79-68.ph.ph.cox.net [68.110.79.68]) by internetstaff.com (Postfix) with ESMTP id 61261F6C0A for ; Mon, 15 Sep 2003 07:38:18 -0700 (MST) Subject: Re: [PERFORM] best arrangement of 3 disks for (insert) From: Cott Lang To: pgsql-general@postgresql.org In-Reply-To: <1063613110.11739.1190.camel@haggis> References: <200309121126.29756.rj@last.fm> <60fzj2t4g6.fsf@dev6.int.libertyrms.info> <1063514688.27300.11.camel@blackbox> <1063613110.11739.1190.camel@haggis> Content-Type: text/plain Message-Id: <1063636871.17237.28.camel@blackbox> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.4 Date: Mon, 15 Sep 2003 07:41:11 -0700 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at postgresql.org X-Archive-Number: 200309/969 X-Sequence-Number: 49189 On Mon, 2003-09-15 at 01:05, Ron Johnson wrote: > Interesting. Where did you put the OS, and what kind of, and how > many, SCSI controllers did you have? I ended up with the OS on the same volume, since it didn't seem to make any difference. I'm using a SuperMicro 6023P chassis with an Adaptec 2010S ZCR controller (64bit/66mhz). I wouldn't recommend SuperMicro to anyone else at this point because instead of hooking up both U320 channels to the 6 drive backplane, they only hook up one. Half the bandwidth, no redundancy. I already had a burp on the SCSI channel during a single drive death take out one box. :( > (If the WAL ever becomes the vehicle for PITR, then it will have > to be on a separate disk [and preferably a separate controller], > even if it slows performance.) Well, it won't have to... but it's certainly a good idea. :) If we ever get PITR, I'll be so happy I won't mind rebuilding my boxes, and hopefully I'll have a better budget at that point. ;^) BTW, I didn't get WORSE performance with the WALs on separate disks, it just wasn't any better. Unfortunately I lost the spreadsheet I had all my results in, so I can't be any more specific.