From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: with ECARTIS (v1.0.0; list xfs); Tue, 05 Feb 2008 06:56:17 -0800 (PST) Received: from cuda.sgi.com (cuda2.sgi.com [192.48.168.29]) by oss.sgi.com (8.12.11.20060308/8.12.11/SuSE Linux 0.7) with ESMTP id m15EuCqB001392 for ; Tue, 5 Feb 2008 06:56:15 -0800 Received: from sandeen.net (localhost [127.0.0.1]) by cuda.sgi.com (Spam Firewall) with ESMTP id 3D1905A0487 for ; Tue, 5 Feb 2008 06:56:34 -0800 (PST) Received: from sandeen.net (sandeen.net [209.173.210.139]) by cuda.sgi.com with ESMTP id zdsWFWoL3N6y2KJx for ; Tue, 05 Feb 2008 06:56:34 -0800 (PST) Message-ID: <47A87901.7010508@sandeen.net> Date: Tue, 05 Feb 2008 08:56:01 -0600 From: Eric Sandeen MIME-Version: 1.0 Subject: Re: Linux mkfs.xfs parameters for large files References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: xfs-bounce@oss.sgi.com Errors-to: xfs-bounce@oss.sgi.com List-Id: xfs To: Jamie Tufnell Cc: xfs@oss.sgi.com Jamie Tufnell wrote: > Hi, > > I've done some research into how to best-prepare our Linux (2.6.18) > XFS filesystem for a read-heavy workload of mostly large files (files > between 50MB and 500MB) and I have some questions on block-sizes / > extents. > > My experience with other filesystems leads me to believe I should > experiment with large block-sizes for this application. In trying to > do so, I found that the Linux XFS implementation forces the block size > to not exceed the page size -- 4kB -- which seems a bit small in this > case. At first, I thought this was *really* bad news for me but I've > since read some posts and the impression I'm getting is it's not that > big of a deal? In your careful benchmarking & testing, which aspects of the performance you've found does not meet your needs or expectations? That's always a good place to start. -Eric