From owner-freebsd-stable@FreeBSD.ORG Sun Mar 30 15:48:48 2003 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 5A8C937B401 for ; Sun, 30 Mar 2003 15:48:48 -0800 (PST) Received: from aurora.peterson.ath.cx (12-254-245-65.client.attbi.com [12.254.245.65]) by mx1.FreeBSD.org (Postfix) with ESMTP id E3F9F43FAF for ; Sun, 30 Mar 2003 15:48:47 -0800 (PST) (envelope-from jlp@peterson.ath.cx) Received: from peterson.ath.cx (localhost [127.0.0.1]) by aurora.peterson.ath.cx (Postfix) with ESMTP id 44B8DC5303 for ; Sun, 30 Mar 2003 16:49:23 -0700 (MST) X-Mailer: exmh version 2.6.2 03/21/2003 with nmh-1.0.4 To: freebsd-stable@freebsd.org In-Reply-To: Message from Chuck Swiger of "Sun, 30 Mar 2003 12:01:17 EST." <3E8722DD.5050703@mac.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Sun, 30 Mar 2003 16:49:23 -0700 From: "Jan L. Peterson" Message-Id: <20030330234923.44B8DC5303@aurora.peterson.ath.cx> Subject: Re: vinum performance X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 30 Mar 2003 23:48:49 -0000 Just to shed some real-world light on this vinum benchmark discussion. At a previous employer we had a few HP systems with hardware RAID boards on them (AMI MegaRaid compatible boards, but some HP part number... I can't remember exactly which). On one system, which was in support of our web site (it stored thousands of small files containing html text that would be referenced in pretty much random order), we configured an HP RS12 (JBOD with 12 disks in it) on the AMI controller as a RAID 5 volume. Performance was excellent (this disk supported mostly reads... once a collection of files had been uploaded, it was not modified). Much better then with the files on local disk, even. On another system, where these files were created before being uploaded, performance with RAID 5 was abysmal. We ended up yanking the AMI controller and using the on-board Symbios controller and using vinum as RAID 0+1 (stripe each group of six disks, mirror one group to the other group, each group on a separate physical controller -- the Sym board had dual channels). With this setup, performance was much better. So the upshot is, if you're doing a lot of writes, don't use RAID 5. If you're doing a lot of reads, and you have hardware RAID 5 available, it's probably a win. -jan- -- Jan L. Peterson If your company is looking for a SAGE level IV, let me know.