From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from cuda.sgi.com (cuda2.sgi.com [192.48.176.25]) by oss.sgi.com (8.14.3/8.14.3/SuSE Linux 0.8) with ESMTP id o9QN8rqP112738 for ; Tue, 26 Oct 2010 18:08:53 -0500 Received: from mail.internode.on.net (localhost [127.0.0.1]) by cuda.sgi.com (Spam Firewall) with ESMTP id 4C472128165 for ; Tue, 26 Oct 2010 16:10:10 -0700 (PDT) Received: from mail.internode.on.net (bld-mail13.adl6.internode.on.net [150.101.137.98]) by cuda.sgi.com with ESMTP id C02FVa8sFBY8taZU for ; Tue, 26 Oct 2010 16:10:10 -0700 (PDT) Date: Wed, 27 Oct 2010 10:10:07 +1100 From: Dave Chinner Subject: Re: XFS Performance on NetApp Message-ID: <20101026231007.GA32255@dastard> References: <201010270033.35894@zmi.at> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <201010270033.35894@zmi.at> List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: xfs-bounces@oss.sgi.com Errors-To: xfs-bounces@oss.sgi.com To: Michael Monnerie Cc: xfs@oss.sgi.com On Wed, Oct 27, 2010 at 12:33:35AM +0200, Michael Monnerie wrote: > Does anybody have a report how XFS behaves on a NetApp storage with thin > provisioning? They have a completely "weird" storage, their WAFL (write > anywhere file layout) together with deduplication and other things make > me think the best is to not-at-all specify any "performance options" in > mkfs/mount, like sunit/swidth or largeio, swalloc, etc. > > Does someone have information on that? Not about the Netapp as such. However, as a general rule thin provisioned storage will have unpredictable performance characteristics as the block number/physical location correlation is meaningless. e.g. because the log is one of the first things written to a thinly provisioned volume during mkfs, it is unlikely to be physically located in the middle of the volume. Indeed, there's no guarantee that it will even be on the same spindles as the rest of the filesystem, and it may be sharing spindles with some other thin volume.... So, you are right in assuming that it is best not to tune your filesystem to a specific physical geometry because there generally isn't one for thinly provisioned volumes. However, options that reduce filesystem fragmentation (e.g. allocsize) still have value in keeping the amount of metadata and ptotential seeks down... Cheers, Dave. -- Dave Chinner david@fromorbit.com _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs