From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755796Ab1IBUkw (ORCPT ); Fri, 2 Sep 2011 16:40:52 -0400 Received: from 87-104-106-3-dynamic-customer.profibernet.dk ([87.104.106.3]:46017 "EHLO kernel.dk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755334Ab1IBUku (ORCPT ); Fri, 2 Sep 2011 16:40:50 -0400 Message-ID: <4E613F4B.8030502@kernel.dk> Date: Fri, 02 Sep 2011 14:40:43 -0600 From: Jens Axboe MIME-Version: 1.0 To: Maxim Patlasov CC: linux-kernel@vger.kernel.org Subject: Re: [PATCH v2] cfq: queued/in_driver stats References: <1314982962-30271-1-git-send-email-maxim.patlasov@gmail.com> In-Reply-To: <1314982962-30271-1-git-send-email-maxim.patlasov@gmail.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 2011-09-02 11:02, Maxim Patlasov wrote: > The patch implements gathering and exporting statistics for cfqd->rq_queued > and cfqd->rq_in_driver. Average values of queued/in_driver are visible via > /sys/.../iosched/queued_avg and /sys/.../iosched/in_driver_avg correspondingly. > The output is in form of "%lu.%02lu %lu.%02lu %lu.%02lu" where the first > value represents average over 1 sec interval, the second - over 5 sec, > the third - over 15 sec. > > The patch also exports cfqd->hw_tag to be seen via /sys/.../iosched/hw_tag. > > Seeing hw_tag is useful to be sure that cfq detected NCQ correctly on given > h/w. queued/in_driver stats are useful to debug performance problems: e.g., > if you know that fast h/w raid is able to concurrently process many requests > quite effectively, but at the same time you observe high queued average and > low in_driver one, you can suggest that cfq underloads h/w raid. > > Changes since first version: do_div was used wrongly, fixed. Seems like an awful lot of code for something that could be gleaned with blktrace? -- Jens Axboe