Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755675AbZCCFgP (ORCPT ); Tue, 3 Mar 2009 00:36:15 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750855AbZCCFf7 (ORCPT ); Tue, 3 Mar 2009 00:35:59 -0500 Received: from lemon.ertos.nicta.com.au ([203.143.174.143]:40699 "EHLO lemon.gelato.unsw.edu.au" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750776AbZCCFf6 (ORCPT ); Tue, 3 Mar 2009 00:35:58 -0500 X-Greylist: delayed 2248 seconds by postgrey-1.27 at vger.kernel.org; Tue, 03 Mar 2009 00:35:58 EST Message-ID: <49ACB8EF.4000600@cse.unsw.edu.au> Date: Tue, 03 Mar 2009 15:58:23 +1100 From: Aaron Carroll User-Agent: Thunderbird 2.0.0.19 (X11/20090213) MIME-Version: 1.0 To: =?UTF-8?B?6LCi57qy?= CC: linux-kernel@vger.kernel.org, Jens Axboe Subject: Re: The difference of request dir between AS and Deadline I/O scheduler? References: <49ACB02D.8070300@cse.unsw.edu.au> In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit X-SA-Do-Not-Run: Yes X-SA-Exim-Connect-IP: 203.143.161.65 X-SA-Exim-Mail-From: aaronc@cse.unsw.edu.au X-SA-Exim-Scanned: No (on lemon.gelato.unsw.edu.au); SAEximRunCond expanded to false Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1395 Lines: 35 谢纲 wrote: > On Tue, Mar 3, 2009 at 12:21 PM, Aaron Carroll wrote: >> � wrote: >>> Hi, >>> >>> I'm little confused about the defination of request dir in AS and >>> Deadline I/O scheduler. >>> In AS, the request dir is defined by wheher it's sync: >>> >>> data_dir = rq_is_sync(rq); >>> >>> But in Deadline, the requests are grouped by read and write. >>> >>> Why is there the difference since AS is an extension of Deadline? >>> what's the consideration? >> I also thought it was silly to have different behaviours, so I tried >> the following patch that makes deadline use sync/async instead of >> read/write. All the benchmarks I tried showed that performance >> dropped or remained constant at best, so I didn't propose it. >> Maybe you will have more luck... > Hello, > Which benchmark tool do you use? I'd like to have a try. I think the > I/O behavior is an important factor which can affect the performance. I can't find the original results, but from memory I tried FIO (many random readers with various I/O sizes), postmark, and compilebench on a single-disk and 10-disk hardware RAID system. -- Aaron -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/