From mboxrd@z Thu Jan 1 00:00:00 1970 From: Theodore Ts'o Subject: Re: [PATCH v2] ext4: dynamical adjust the length of zero-out chunk Date: Mon, 13 Aug 2012 14:40:53 -0400 Message-ID: <20120813184053.GE32484@thunk.org> References: <1342075703-4537-1-git-send-email-wenqing.lz@taobao.com> <4FFEE402.7090601@redhat.com> <7B794B69-EF6C-4279-83D7-EA47E35CD54C@dilger.ca> <20120813032243.GB13072@thunk.org> <20120813173224.GF9117@lenny.home.zabbo.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Andreas Dilger , Zheng Liu , Eric Sandeen , Zheng Liu , ext4 development To: Zach Brown Return-path: Received: from li9-11.members.linode.com ([67.18.176.11]:44470 "EHLO imap.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752113Ab2HMSlE (ORCPT ); Mon, 13 Aug 2012 14:41:04 -0400 Content-Disposition: inline In-Reply-To: <20120813173224.GF9117@lenny.home.zabbo.net> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Mon, Aug 13, 2012 at 10:32:24AM -0700, Zach Brown wrote: > > we allow it to be tunable via sysfs and set an initial default value > > of 32, so instead of creating uninitalized extents smaller than > > s/32/16/? Oops, nice catch. > It'd be nice to define the tunable in terms of some fixed unit, kb or > mb, whatever, and then translate to the block size in the code so people > don't have to do that math by hand. > > No? Agreed, thanks for the suggestion. The next question is whether the default maximum zero-out size should be 256k (as previously documented) or 128k (as previously coded). The previously rule of thumb which I had used was that after doing a random seek, the time needed to write 4k and 32k was pretty much in the noise. But that was a number from several years ago. I suppose I should do some quick experiments to see what is a good number these days.... - Ted