From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from relay.sgi.com (relay3.corp.sgi.com [198.149.34.15]) by oss.sgi.com (Postfix) with ESMTP id 215E57CA0 for ; Tue, 9 Aug 2016 16:26:23 -0500 (CDT) Received: from cuda.sgi.com (cuda1.sgi.com [192.48.157.11]) by relay3.corp.sgi.com (Postfix) with ESMTP id A6446AC001 for ; Tue, 9 Aug 2016 14:26:22 -0700 (PDT) Received: from ipmail04.adl6.internode.on.net (ipmail04.adl6.internode.on.net [150.101.137.141]) by cuda.sgi.com with ESMTP id ZslnHApBFRrAFalI for ; Tue, 09 Aug 2016 14:26:11 -0700 (PDT) Date: Wed, 10 Aug 2016 07:26:07 +1000 From: Dave Chinner Subject: Re: Advice needed with file system corruption Message-ID: <20160809212607.GH19025@dastard> References: <5787852A.7030900@st-andrews.ac.uk> <20160808161132.1d76eb5c@harpe.intellique.com> <1470670691.8958.11.camel@filmlight.ltd.uk> <20160808174427.24d8272c@harpe.intellique.com> <592673683.10446004.1470715375850.JavaMail.yahoo@mail.yahoo.com> <7faf941c-2188-589d-b624-9fedb22dee40@telefonica.net> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <7faf941c-2188-589d-b624-9fedb22dee40@telefonica.net> 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 Errors-To: xfs-bounces@oss.sgi.com Sender: xfs-bounces@oss.sgi.com To: "Carlos E. R." Cc: XFS mail list On Tue, Aug 09, 2016 at 02:40:26PM +0200, Carlos E. R. wrote: > On 2016-08-09 06:02, Gim Leong Chin wrote: > > > Drives connected to RAID controllers with battery backed cache should > > have their caches "disabled" (they are really set to write through mode > > instead). By the way, I found out in lab testing that 7200 RPM SATA > > drives suffer a big performance loss when doing sequential writes in > > cache write through mode. > > If you disable the disk internal cache, as a consequence you also > disable the disk internal write optimizations. It has to be much slower > at writing. It seems to me obvious. This is why decent HW RAID controllers have a large non volatile write cache - the caching is done in the controller where it is safe from power loss, not in the drive where it is unsafe. Write optimisations happen at the RAID controller level, not at the individual drive level. As for 10/15krpm SAS drive performance, they generally are only slower in microbenchmark situations (e.g. sequential single sector writes) when the write cache is disabled. These sorts of loads aren't typically seen in the real world, so for most people there is little difference in performance on high end enterprise SAS drives when changing the cache mode.... Cheers, Dave. -- Dave Chinner david@fromorbit.com _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs