From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christopher Li Subject: Re: Yet another base64 patch Date: Wed, 13 Apr 2005 22:42:28 -0400 Message-ID: <20050414024228.GC18655@64m.dyndns.org> References: <425DEF64.60108@zytor.com> <20050414022413.GB18655@64m.dyndns.org> <425E0174.4080404@zytor.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: git@vger.kernel.org X-From: git-owner@vger.kernel.org Thu Apr 14 07:49:41 2005 Return-path: Received: from vger.kernel.org ([12.107.209.244]) by ciao.gmane.org with esmtp (Exim 4.43) id 1DLxE7-0002RY-TF for gcvg-git@gmane.org; Thu, 14 Apr 2005 07:49:20 +0200 Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261287AbVDNFw0 (ORCPT ); Thu, 14 Apr 2005 01:52:26 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261429AbVDNFw0 (ORCPT ); Thu, 14 Apr 2005 01:52:26 -0400 Received: from rwcrmhc12.comcast.net ([216.148.227.85]:14819 "EHLO rwcrmhc12.comcast.net") by vger.kernel.org with ESMTP id S261287AbVDNFwV (ORCPT ); Thu, 14 Apr 2005 01:52:21 -0400 Received: from localhost.localdomain (c-24-6-236-77.hsd1.ca.comcast.net[24.6.236.77]) by comcast.net (rwcrmhc12) with ESMTP id <2005041405522001400hbd71e>; Thu, 14 Apr 2005 05:52:20 +0000 Received: by localhost.localdomain (Postfix, from userid 1027) id 55F1C3F1EF; Wed, 13 Apr 2005 22:42:28 -0400 (EDT) To: "H. Peter Anvin" Content-Disposition: inline In-Reply-To: <425E0174.4080404@zytor.com> User-Agent: Mutt/1.4.1i Sender: git-owner@vger.kernel.org Precedence: bulk X-Mailing-List: git@vger.kernel.org On Wed, Apr 13, 2005 at 10:36:52PM -0700, H. Peter Anvin wrote: > Christopher Li wrote: > >On Wed, Apr 13, 2005 at 09:19:48PM -0700, H. Peter Anvin wrote: > > > >That is not surprising due to the directory index in ext3. Htree is pretty > >good at random access and the hashed file name distribute evenly, that is > >the best case for htree. > > > > Right, so by not trying to do the filesystem's job for it we actually > come out ahead. > But if you write a large number of random files, when htree has three levels index. htree will suffer on the effect that it dirty random block very quickly, most block get dirty only contain one or two new entries. Ext3 will choke on it due to the limited journal size. While non-index directory, new entry are very compact on the blocks. So it end up dirty a lot less blocks, of course, lookup will suffer. Depend on you want check out fast or write a big tree fast, you can't win it all. Chris