From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261947AbUCON1L (ORCPT ); Mon, 15 Mar 2004 08:27:11 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S262571AbUCON1L (ORCPT ); Mon, 15 Mar 2004 08:27:11 -0500 Received: from ns.suse.de ([195.135.220.2]:38545 "EHLO Cantor.suse.de") by vger.kernel.org with ESMTP id S261947AbUCON06 (ORCPT ); Mon, 15 Mar 2004 08:26:58 -0500 Subject: Re: aio tiobench From: Chris Mason To: William Lee Irwin III Cc: linux-aio@kvack.org, linux-kernel@vger.kernel.org In-Reply-To: <20040315080520.GC655@holomorphy.com> References: <20040315080520.GC655@holomorphy.com> Content-Type: text/plain Message-Id: <1079357322.4186.377.camel@watt.suse.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.5 Date: Mon, 15 Mar 2004 08:28:43 -0500 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Mon, 2004-03-15 at 03:05, William Lee Irwin III wrote: > So I farted around for a hour or two seeing if I could get tiobench > to do aio for the general purpose of exercising codepaths, benchmarking, > etc. in simple ways. Hopefully this answers the need for regular, > simple, and easily-available methods of exercising and/or benchmarking > the aio code in some way. You might want to check out the list of benchmarks collected at: http://lse.sourceforge.net/io/aio.html I found that adapting existing benchmarks made it hard to test some of the aio corner cases, so aio-stress is just a big state machine, with options to tweak how badly you want to abuse the aio subsystem. There are a few other good ones on the page though. -chris