From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 15 May 2001 11:44:02 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 15 May 2001 11:43:52 -0400 Received: from roc-24-169-102-121.rochester.rr.com ([24.169.102.121]:44813 "EHLO roc-24-169-102-121.rochester.rr.com") by vger.kernel.org with ESMTP id ; Tue, 15 May 2001 11:43:37 -0400 Date: Tue, 15 May 2001 11:42:16 -0400 From: Chris Mason To: Ricardo Galli , linux-kernel@vger.kernel.org Subject: Re: Reiserfs, Mongo and CPU question Message-ID: <1076120000.989941336@tiny> In-Reply-To: X-Mailer: Mulberry/2.0.8 (Linux/x86) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Tuesday, May 15, 2001 01:41:01 PM +0200 Ricardo Galli wrote: > Hans and reiserfs developers, > the same student of my university > (http://www.cs.helsinki.fi/linux/linux-kernel/2001-18/0654.html) was > carrying up the mongo benchmarks against reiser, xfs, jfs and ext2 for > different base sizes. > > > For example, for the base size of 10.000 (the average of a clean > distribution is about 16.000 bytes) ReiserFS is even slower than ext2. > I've realised the bottleneck may be the CPU, a Cyrix MII 233MHz. > Would not surprise me, there's lots of room for improvement in reiserfs CPU usage. The 10k size is one of the worst cases for tail performance, those numbers should increase if you mount with -o notail. Here's a simple patch that should help on balance instensive apps (like creates/deletes). Please let me know if you see any difference with it. -chris diff -ur diff/linux/fs/reiserfs/fix_node.c linux/fs/reiserfs/fix_node.c --- diff/linux/fs/reiserfs/fix_node.c Mon Jan 15 18:31:19 2001 +++ linux/fs/reiserfs/fix_node.c Fri Feb 2 15:40:54 2001 @@ -936,6 +936,7 @@ if (p_s_tb->FEB[p_s_tb->cur_blknum]) BUG(); + mark_buffer_journal_new(p_s_new_bh) ; p_s_tb->FEB[p_s_tb->cur_blknum++] = p_s_new_bh; }