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 324CF3A5834 for ; Wed, 26 Jan 2005 21:16:41 +0000 (GMT) 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 58726-07 for ; Wed, 26 Jan 2005 21:16:37 +0000 (GMT) Received: from mail.tecarta.com (66.238.115.135.ptr.us.xo.net [66.238.115.135]) by svr1.postgresql.org (Postfix) with ESMTP id 945DB3A581E for ; Wed, 26 Jan 2005 21:16:37 +0000 (GMT) Received: from mail pickup service by mail.tecarta.com with Microsoft SMTPSVC; Wed, 26 Jan 2005 13:19:11 -0800 Received: from mail.tecarta.com ([192.168.160.2]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Wed, 26 Jan 2005 13:19:09 -0800 Received: from barracuda.tecarta.com ([192.168.160.200]) by mail.tecarta.com (SAVSMTP 3.1.0.29) with SMTP id M2005012613190901130 for ; Wed, 26 Jan 2005 13:19:09 -0800 X-ASG-Debug-ID: 1106774194-14021-5-0 X-Barracuda-URL: http://192.168.160.200:8000/cgi-bin/mark.cgi Received: from mail2 (mail2.hq.corp [192.168.160.6]) by barracuda.tecarta.com (Spam Firewall) with SMTP id 48AE320147D1 for ; Wed, 26 Jan 2005 13:16:34 -0800 (PST) Received: from [192.168.0.13] ([63.206.203.145]) by mail.tecarta.com with Microsoft SMTPSVC(6.0.3790.0); Wed, 26 Jan 2005 13:19:04 -0800 Message-ID: <41F79610.20003@sfnet.cc> Date: Wed, 26 Jan 2005 13:07:28 +0000 From: Steve Poe User-Agent: Mozilla Thunderbird 1.0 (X11/20041228) X-Accept-Language: en-us, en MIME-Version: 1.0 To: Josh Berkus Cc: pgsql-performance@postgresql.org X-ASG-Orig-Subj: Re: [PERFORM] Ideal disk setup for Postgresql 7.4? Subject: Re: Ideal disk setup for Postgresql 7.4? References: <41F63424.1060904@sfnet.cc> <200501251903.09734.josh@agliodbs.com> <41F6DBD8.2050008@sfnet.cc> <200501261003.15267.josh@agliodbs.com> In-Reply-To: <200501261003.15267.josh@agliodbs.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 26 Jan 2005 21:19:04.0536 (UTC) FILETIME=[A9C51D80:01C503EC] X-Virus-Scanned: by Barracuda Spam Firewall at tecarta.com X-Barracuda-Spam-Score: -0.06 X-Barracuda-Spam-Status: No, SCORE=-0.06 using global scores of TAG_LEVEL=4.0 QUARANTINE_LEVEL=1000.0 KILL_LEVEL=7.0 tests=BAYES_30, DATE_IN_PAST_06_12 X-Barracuda-Spam-Report: Code version 2.64, rules version 2.1.960 Rule breakdown below pts rule name description ---- ---------------------- ------------------------------------------- -0.90 BAYES_30 BODY: Bayesian spam probability is 30 to 40% [score: 0.3659] 0.85 DATE_IN_PAST_06_12 Date: is 6 to 12 hours before Received: date X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.269 tagged_above=0 required=5 tests=AWL, DATE_IN_PAST_06_12, FORGED_RCVD_HELO X-Spam-Level: X-Archive-Number: 200501/516 X-Sequence-Number: 10206 >FWIW, 7.4.6 is a binary, drop-in place upgrade for 7.4.2. And 7.4.2 has known >bugs. However, I understand your situation. > > > As soon as we get the go-ahead, I will upgrade. I think the company is actually looking towards 8.0 certification. >>Okay, thanks. Even with 7-disks? I trust that. >> >> > >Well, it's less bad with 7 disks than it is with 3, certainly. However,there >is an obvious and quick gain to be had by splitting off the WAL logs onto >their own disk resource ... up to 14%+ performance in some applications. > > > Pardon my ignorance, but the WAL logs are comprised of pg_xlog and pg_clog? Their own disk resource, but not within the same channel of disks the database is on, right? >>So, RAID 1+0 (sw) is >>probably the best option. I've run sw RAID personally for years without >>issue. I am a bit hesitant in doing sw RAID for a production server for >>a database --- probably because its not my server. Any thoughts on sw >>RAID for Postgresql? >> >> > >Yes. See my article for one. In generaly, SW RAID on BSD or Linux works >well for PostgreSQL ... UNLESS your machine is already CPU-bound, in which >case it's a bad idea. If you're hitting the CS bug, it's definitely a bad >idea, because the SW RAID will increase context switching. > >So if your choice, on your system, is between sw RAID 10, and hw RAID 5, and >you're having excessive CSes, I'd stick with the HW RAID. > > > Okay. InCPU-bound servers, use hw RAID. Any hw raids to avoid? >>Okay. Darn. While I don't write the queries for the application, I do >>interact with the company frequently. Their considering moving the >>queries into the database with PL/pgSQL. Currently their queries are >>done through ProvIV development using ODBC. Will context switching be >>minimized here by using PL/pgSQL? >> >> > >Won't make a difference, actually. Should improve performance in other ways, >though, by reducing round-trip time on procedures. Feel free to recommend >the company to this list. > > > I think their too busy to monitor/watch this list. Not a put-down to them, but I have to do my own leg work to help decide what we're going to do. >>Dual Xeon 2.8 CPUs with HT turned off. >> >> > >Yeah, thought it was a Xeon. > > > If we went with a single CPU, like Athlon/Opertron64, would CS storming go away? Thanks. Steve Poe