From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S265394AbUEUHvR (ORCPT ); Fri, 21 May 2004 03:51:17 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S265426AbUEUHvR (ORCPT ); Fri, 21 May 2004 03:51:17 -0400 Received: from ns.virtualhost.dk ([195.184.98.160]:17621 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S265394AbUEUHvP (ORCPT ); Fri, 21 May 2004 03:51:15 -0400 Date: Fri, 21 May 2004 09:50:28 +0200 From: Jens Axboe To: Nick Piggin Cc: Andrew Morton , alexeyk@mysql.com, linuxram@us.ibm.com, peter@mysql.com, linux-kernel@vger.kernel.org Subject: Re: Random file I/O regressions in 2.6 [patch+results] Message-ID: <20040521075027.GN1952@suse.de> References: <200405022357.59415.alexeyk@mysql.com> <1084480888.22208.26.camel@dyn319386.beaverton.ibm.com> <1084815010.13559.3.camel@localhost.localdomain> <200405200506.03006.alexeyk@mysql.com> <20040520145902.27647dee.akpm@osdl.org> <20040520152305.3dbfa00b.akpm@osdl.org> <40ADB062.8050005@yahoo.com.au> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <40ADB062.8050005@yahoo.com.au> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, May 21 2004, Nick Piggin wrote: > Andrew Morton wrote: > > >Open questions are: > > > >a) Why is 2.6 write coalescing so superior to 2.4? > > > >b) Why is 2.6 issuing more read requests, for less data? > > > >c) Why is Alexey seeing dissimilar results? > > > > > Interesting. I am not too familiar with 2.4's IO scheduler, > but 2.6's have pretty comprehensive merging systems. Could > that be helping, Jens? Or is 2.4 pretty equivalent? 2.4 will give up merging faster than 2.6, elevator_linus will stop looking for a merge point if the sequence drops to zero. 2.6 will always merge. So that could explain the fewer writes. > What about things like maximum request size for 2.4 vs 2.6 > for example? This is another thing that can have an impact, > especially for writes. I think that's pretty similar. Andrew didn't say what device he was testing on, but 2.4 ide defaults to max 64k where 2.6 defaults to 128k. > I'll take a guess at b, and say it could be as-iosched.c. > Another thing might be that 2.6 has smaller nr_requests than > 2.4, although you are unlikely to hid the read side limit > with only 16 threads if they are doing sync IO. Andrew, you did numbers for deadline previously as well, but no rq statistics there? As for nr_requests that's true, would be worth a shot to bump available requests in 2.6. -- Jens Axboe