From owner-freebsd-fs@FreeBSD.ORG Tue Nov 1 03:52:58 2011 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id A83C4106564A; Tue, 1 Nov 2011 03:52:58 +0000 (UTC) (envelope-from rmacklem@uoguelph.ca) Received: from esa-annu.mail.uoguelph.ca (esa-annu.mail.uoguelph.ca [131.104.91.36]) by mx1.freebsd.org (Postfix) with ESMTP id 52B8A8FC0C; Tue, 1 Nov 2011 03:52:57 +0000 (UTC) X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: Ap4EAH1sr06DaFvO/2dsb2JhbABDFoRhpUWBcgEpVjIDAg0ZAlEOiB2iTpF8gTCGPoEUBJQPkXs X-IronPort-AV: E=Sophos;i="4.69,436,1315195200"; d="scan'208";a="142153541" Received: from erie.cs.uoguelph.ca (HELO zcs3.mail.uoguelph.ca) ([131.104.91.206]) by esa-annu-pri.mail.uoguelph.ca with ESMTP; 31 Oct 2011 23:52:57 -0400 Received: from zcs3.mail.uoguelph.ca (localhost.localdomain [127.0.0.1]) by zcs3.mail.uoguelph.ca (Postfix) with ESMTP id 7A550B3F0C; Mon, 31 Oct 2011 23:52:57 -0400 (EDT) Date: Mon, 31 Oct 2011 23:52:57 -0400 (EDT) From: Rick Macklem To: FS List Message-ID: <780905615.915177.1320119577445.JavaMail.root@erie.cs.uoguelph.ca> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-Originating-IP: [172.17.91.203] X-Mailer: Zimbra 6.0.10_GA_2692 (ZimbraWebClient - FF3.0 (Win)/6.0.10_GA_2692) Cc: Subject: NFS client patch that implements a list of dirty byte ranges in a buffer X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 01 Nov 2011 03:52:58 -0000 Hi, I created this patch after the discussion by John Baldwin in late August under the Subject: Fixes to allow write clustering of NFS writes from a FreeBSD NFS client First off, I am in no way trying to "steal John's thunder" and I hope he will continue pursuing the good work he's doing on the above. However, I thought that the patch might be interesting to others who like to live on the "bleeding edge" and use NFS for source tree building, since the patch seems to help with the performance of that. My simple test using a run of "make -j4 buildworld" without/with the patch reduced the elapsed time from 4:17 to 2:42. (This was in no way a benchmark and used very slow single core i386 hardware, but it at least suggests that the patch may be useful.) So, if anyone is interested in trying the patch, it is at: http://people.freebsd.org/~rmacklem/dirtybuflist.patch But, be forewarned, this patch has been minimally tested and only on single core i386 hardware. If you do try the patch, please let me know how it goes, rick