From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263775AbTE3QHx (ORCPT ); Fri, 30 May 2003 12:07:53 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263777AbTE3QHx (ORCPT ); Fri, 30 May 2003 12:07:53 -0400 Received: from pao-ex01.pao.digeo.com ([12.47.58.20]:23724 "EHLO pao-ex01.pao.digeo.com") by vger.kernel.org with ESMTP id S263775AbTE3QHv (ORCPT ); Fri, 30 May 2003 12:07:51 -0400 Date: Fri, 30 May 2003 09:21:11 -0700 From: Andrew Morton To: Nick Piggin Cc: jacobs@penguin.theopalgroup.com, linux-kernel@vger.kernel.org Subject: Re: Ext3 meta-data performance Message-Id: <20030530092111.5bdadf5c.akpm@digeo.com> In-Reply-To: <3ED772F5.8060100@cyberone.com.au> References: <3ED772F5.8060100@cyberone.com.au> X-Mailer: Sylpheed version 0.9.0pre1 (GTK+ 1.2.10; i686-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 30 May 2003 16:21:11.0672 (UTC) FILETIME=[7BF83F80:01C326C7] Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Nick Piggin wrote: > > > > >I have some more time to experiment, so please let me know if there is > >anything else you think I should try. > > > Andrew might be able to suggest some worthwhile tests, if nothing > else, try mounting your filesystems as ext2, so you can get a > baseline figure. So the workload is a `cp -Rl' of a 500,000 file tree? Vast amounts of metadata. ext2 will fly through that. Poor old ext3 has to write everything twice, and has to keep on doing seek-intensive checkpointing when the journal fills. When we get Andreas's "dont bother reading empty inode blocks" speedup going it will help both filesystems quite a bit. Increasing the journal size _may_ help. `tune2fs -J size=400', or `mke2fs -j J size=400'. The Orlov allocator will help this workload significantly, but you have to give it time to settle in: it uses different decisions for the placement of directories on disk. If the source directory of the copy was created by a 2.4 filesystem then we won't get any of that benefit.