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 9DD7A3A5851 for ; Tue, 11 Jan 2005 04:04:55 +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 90309-09 for ; Tue, 11 Jan 2005 04:04:41 +0000 (GMT) Received: from floppy.pyrenet.fr (news.pyrenet.fr [194.250.190.2]) by svr1.postgresql.org (Postfix) with ESMTP id B17373A57AD for ; Tue, 11 Jan 2005 04:04:40 +0000 (GMT) Received: by floppy.pyrenet.fr (Postfix, from userid 106) id 41DB13158F; Tue, 11 Jan 2005 05:04:39 +0100 (MET) From: Christopher Browne X-Newsgroups: pgsql.performance Subject: Re: which dual-CPU hardware/OS is fastest for PostgreSQL? Date: Mon, 10 Jan 2005 23:04:20 -0500 Organization: cbbrowne Computing Inc Lines: 61 Message-ID: References: <59b2d39b05011018421680264b@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Complaints-To: usenet@news.hub.org X-message-flag: Outlook is rather hackable, isn't it? X-Home-Page: http://www.cbbrowne.com/info/ X-Affero: http://svcs.affero.net/rm.php?r=cbbrowne User-Agent: Gnus/5.1007 (Gnus v5.10.7) XEmacs/21.4 (Corporate Culture, linux) Cancel-Lock: sha1:Z1W6VL2SgQAOwJhB30+vNJnmqNI= To: pgsql-performance@postgresql.org X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.323 tagged_above=0 required=5 tests=AWL, INFO_TLD X-Spam-Level: X-Archive-Number: 200501/122 X-Sequence-Number: 9812 Quoth mileskeaton@gmail.com (Miles Keaton): > I'm sorry if there's a URL out there answering this, but I couldn't > find it. > > For those of us that need the best performance possible out of a > dedicated dual-CPU PostgreSQL server, what is recommended? > > AMD64/Opteron or i386/Xeon? Xeon sux pretty bad... > Linux or FreeBSD or _?_ The killer question won't be of what OS is "faster," but rather of what OS better supports the fastest hardware you can get your hands on. We tried doing some FreeBSD benchmarking on a quad-Opteron box, only to discover that the fibrechannel controller worked in what amounted to a "PAE-like" mode where it only talked DMA in a 32 bit manner. We might have found a more suitable controller, given time that was not available. A while back I tried to do some FreeBSD benchmarking on a quad-Xeon box with 8GB of RAM. I couldn't find _any_ RAID controller compatible with that configuration, so FreeBSD wasn't usable on that hardware unless I told the box to ignore half the RAM. There lies the rub of the problem: you need to make sure all the vital components are able to run "full blast" in order to maximize performance. The really high end SCSI controllers may only have supported drivers for some specific set of OSes, and it seems to be pretty easy to put together boxes where one or another component leaps into the "That Doesn't Work!" category. > I'm assuming hardware RAID 10 on 15k SCSI drives is fastest disk > performance. RAID controllers tend to use i960 or StrongARM CPUs that run at speeds that _aren't_ all that impressive. With software RAID, you can take advantage of the _enormous_ increases in the speed of the main CPU. I don't know so much about FreeBSD's handling of this, but on Linux, there's pretty strong indication that _SOFTWARE_ RAID is faster than hardware RAID. It has the further merit that you're not dependent on some disk formatting scheme that is only compatible with the model of RAID controller that you've got, where if the controller breaks down, you likely have to rebuild the whole array from scratch and your data is toast. The assumptions change if you're looking at really high end disk arrays, but that's certainly another story. -- (format nil "~S@~S" "cbbrowne" "acm.org") http://linuxfinances.info/info/finances.html Real Programmers are surprised when the odometers in their cars don't turn from 99999 to A0000.