From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263425AbVGAST5 (ORCPT ); Fri, 1 Jul 2005 14:19:57 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263423AbVGAST5 (ORCPT ); Fri, 1 Jul 2005 14:19:57 -0400 Received: from ns.virtualhost.dk ([195.184.98.160]:1990 "EHLO virtualhost.dk") by vger.kernel.org with ESMTP id S263420AbVGASTx (ORCPT ); Fri, 1 Jul 2005 14:19:53 -0400 Date: Fri, 1 Jul 2005 20:21:12 +0200 From: Jens Axboe To: "Miller, Mike (OS Dev)" Cc: akpm@osdl.org, linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org Subject: Re: [PATCH 1/1] cciss per disk queue for 2.6 Message-ID: <20050701182111.GA24146@suse.de> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Jul 01 2005, Miller, Mike (OS Dev) wrote: > Jens Axboe wrote: > > Different thing, I'm talking about single volume starvation, > > not volume-to-volume. > > > > > elevator algoritm(s) may be causing writes to starve reads > > on the same > > > logical volume. We continue to investigate our other > > performance issues. > > > > I completely disagree. Even with an intelligent io scheduler, > > starvation is seen on ciss that does not happen on other > > queueing hardware such as 'normal' scsi controllers/drives. > > So something else is going on, and the only 'fix' so far is > > to limit the ciss queue depth heavily. > > We will investigate this further and come up with a solution. Could be > the firmware, I suppose. Unfortunately I don't have any hardware at hand for testing, so I cannot do it myself. Would be appreciated! -- Jens Axboe