From owner-freebsd-questions@FreeBSD.ORG Sun Jun 26 07:42:00 2011 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 0806C1065677 for ; Sun, 26 Jun 2011 07:42:00 +0000 (UTC) (envelope-from dan@dan.emsphone.com) Received: from email2.allantgroup.com (email2.emsphone.com [199.67.51.116]) by mx1.freebsd.org (Postfix) with ESMTP id A70EA8FC25 for ; Sun, 26 Jun 2011 07:41:59 +0000 (UTC) Received: from dan.emsphone.com (dan.emsphone.com [199.67.51.101]) by email2.allantgroup.com (8.14.4/8.14.4) with ESMTP id p5Q7fwoQ086908 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Sun, 26 Jun 2011 02:41:58 -0500 (CDT) (envelope-from dan@dan.emsphone.com) Received: from dan.emsphone.com (smmsp@localhost [127.0.0.1]) by dan.emsphone.com (8.14.5/8.14.5) with ESMTP id p5Q7fmWS032523 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Sun, 26 Jun 2011 02:41:48 -0500 (CDT) (envelope-from dan@dan.emsphone.com) Received: (from dan@localhost) by dan.emsphone.com (8.14.5/8.14.5/Submit) id p5Q7fjAW032492; Sun, 26 Jun 2011 02:41:45 -0500 (CDT) (envelope-from dan) Date: Sun, 26 Jun 2011 02:41:45 -0500 From: Dan Nelson To: Joshua Isom Message-ID: <20110626074141.GB44024@dan.emsphone.com> References: <4E06180F.7090409@gmail.com> <16AE025C-EA55-45F3-8419-542228CE5AA4@my.gd> <4E06C2D1.4050601@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4E06C2D1.4050601@gmail.com> X-OS: FreeBSD 8.2-STABLE User-Agent: Mutt/1.5.21 (2010-09-15) X-Virus-Scanned: clamav-milter 0.97 at email2.allantgroup.com X-Virus-Status: Clean X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.2.6 (email2.allantgroup.com [199.67.51.78]); Sun, 26 Jun 2011 02:41:58 -0500 (CDT) X-Scanned-By: MIMEDefang 2.68 on 199.67.51.78 Cc: Damien Fleuriot , "freebsd-questions@freebsd.org" Subject: Re: Performance of a USB ZIL for ZFS X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 26 Jun 2011 07:42:00 -0000 In the last episode (Jun 26), Joshua Isom said: > On 6/25/2011 9:32 PM, Damien Fleuriot wrote: > > On 25 Jun 2011, at 19:17, Joshua Isom wrote: > >> I was wondering if anyone had tried using a decent USB flash drive for > >> the ZIL. I know it'd be hard finding one fast enough, but some from > >> patriot seem like they might be suitable for home use. Part of the > >> idea is to just minimize hard drive thrashing and the wear and tear > >> associated with it. If it helps prevent the drives from going bad, and > >> doesn't hurt performance too bad all the better. But if it's going to > >> hurt performance too much or not help prevent thrashing there isn't a > >> point. > > > > I stopped reading at the title. > > The answer is no. > > > > Grab a SSD for $80-120ish. > > Perhaps it would have helped to read the email. Part of the concern is > making sure the drives don't fail and not just throughput. > > Given that Kingston sells an SATA SSD for $40 that only gets writes at > 30mb/s write, and some USB drives might get up to 20mb/s. If I get two > drives and put them on different controllers, mirrored, I might get > acceptable performance. I may still loose performance, but if my drives > last a year longer, I can probably accept it. I'm ok with loosing some > performance, but I just don't want it dragging down the system. And if it > won't help the drives last longer there's no point. A seaparate ZIL isn't meant to extend the lifetime of the hard drives; it's meant to accelerate the speed of sync writes. Those are pretty infrequent themselves, unless you're an NFS server. You'll see a couple syncs per commit on a database server, but compared to the amount of regular reads and writes on your average system, you'll save under 1% of the writes by adding a fast ZIL. And remember, the ZIL is just a write log. Everything that gets written to it will get flushed to disk when zfs writes the next transaction group. -- Dan Nelson dnelson@allantgroup.com