From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stan Hoeppner Subject: Re: What layout with 6x2TB, Intel Atom? Date: Tue, 05 Feb 2013 08:17:09 -0600 Message-ID: <51111465.8080603@hardwarefreak.com> References: <733670.23.1360022484693.JavaMail.root@zimbra> Reply-To: stan@hardwarefreak.com Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <733670.23.1360022484693.JavaMail.root@zimbra> Sender: linux-raid-owner@vger.kernel.org To: Roy Sigurd Karlsbakk Cc: =?ISO-8859-1?Q?Mathias_Bur=E9n?= , Linux RAID List-Id: linux-raid.ids On 2/4/2013 6:01 PM, Roy Sigurd Karlsbakk wrote: >> Third, use XFS, and format with alignment to the md device: >> ~$ mkfs.xfs -d su=512k,sw=4 > > Why? ext4 should be just as good for that, XFS generally yields better performance with concurrency, and is simply a better filesystem across the board WRT performance, features, and tools. And XFS has been designed and optimized since 1994 specifically for storing and streaming large files, making it a perfect fit for the OP's 4GB files. Why use "just as good" when one can use the best? > and better for metadata XFS metadata performance has exceeded that of EXT4 since mainline kernel 2.6.39 (May 2011) for workloads with medium to large concurrency (2-4 threads or more). It's single thread metadata performance is now not far behind EXT4. Thus journal metadata performance is no longer a reason to avoid XFS. Regardless, metadata performance is irrelevant to the OP's stated workload. -- Stan