From owner-freebsd-geom@FreeBSD.ORG Sun Feb 26 23:20:29 2006 Return-Path: X-Original-To: freebsd-geom@freebsd.org Delivered-To: freebsd-geom@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 2EAE816A420 for ; Sun, 26 Feb 2006 23:20:29 +0000 (GMT) (envelope-from terje+geom@elde.net) Received: from smtp.elde.net (us.elde.net [66.246.223.215]) by mx1.FreeBSD.org (Postfix) with ESMTP id 9CE9243D53 for ; Sun, 26 Feb 2006 23:20:28 +0000 (GMT) (envelope-from terje+geom@elde.net) Received: from smtp.elde.net (343006.ds.nac.net [127.0.0.1]) by smtp.elde.net (Postfix) with ESMTP id 84A9365003; Mon, 27 Feb 2006 00:20:27 +0100 (CET) Received: from [127.0.0.1] (60.80-203-96.nextgentel.com [80.203.96.60]) (Authenticated sender: terje@elde.net) by smtp.elde.net (Postfix) with ESMTP id D4F8664F56; Mon, 27 Feb 2006 00:20:26 +0100 (CET) Message-ID: <44023791.2090008@elde.net> Date: Mon, 27 Feb 2006 00:19:45 +0100 From: Terje Elde User-Agent: Thunderbird 1.5 (Windows/20051201) MIME-Version: 1.0 To: OxY References: <000b01c63b0f$c0779b10$0201a8c0@oxy> In-Reply-To: <000b01c63b0f$c0779b10$0201a8c0@oxy> Content-Type: text/plain; charset=ISO-8859-2; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: ClamAV using ClamSMTP Cc: freebsd-geom@freebsd.org Subject: Re: which mirror balance algorithm to use? X-BeenThere: freebsd-geom@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: GEOM-specific discussions and implementations List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 26 Feb 2006 23:20:29 -0000 OxY wrote: > my box is a file server, it has about ~100-150 simultaneous connection > all the time, so the disks are very busy.. > i'd like to use the best performance balance algorithm to reduce disk > load, which is about 80-90% now.. > which one should i use? 'load' and 'round-robin'? > why? > There are many factors here. If you're using both slow and fast disks, using the load algorithm might have some advantages for example. I know this isn't the answer you want, but the best way to go about this might actually be to test the different algorithms. It's easy enough to change, and then you can run up some numbers on how well it seem to be performing with the different ones in your setup. Best thing would be if you could run with a simulated load after hours, so you can get the exact same test patterns run for the different algorithms. Combine with trying them out with real world load, and things should get interesting. Terje