From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx2.fusionio.com ([64.244.102.31]:51804 "EHLO mx2.fusionio.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756897Ab1COU2U (ORCPT ); Tue, 15 Mar 2011 16:28:20 -0400 Message-ID: <4D7FCBDE.2080206@fusionio.com> Date: Tue, 15 Mar 2011 21:28:14 +0100 From: Jens Axboe MIME-Version: 1.0 Subject: Re: memory caching References: In-Reply-To: Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Sender: fio-owner@vger.kernel.org List-Id: fio@vger.kernel.org To: "N. Harake" Cc: "fio@vger.kernel.org" On 2011-03-15 13:36, N. Harake wrote: >=20 > I noticed when I am using --sync=3D1, IOPS go down to 15K and throughput = to > 53MB (That=EF=BF=BDs horrible for the HW I am using), removing --sync I g= ot 150K > IOPS and almost 800MB, which make more sense. What does --sync does, and > hope what I am doing should be enough to avoid any kind of memory caching= .. >=20 > Regards > H. N. Harake >=20 > fio --name=3Djob --directory=3D/lvm --size=3D1G --iodepth=3D1 --sync=3D1 > --filesize=3D1G --direct=3D1 --numjobs=3D64 --bs=3D4096 --rw=3Drandwrite > --group_reporting If your goal is to maximize iops for testing, use the raw block device. Use direct=3D1, and ioengine=3Dlibaio. Set depth large, ala: --iodepth=3D128 --iodepth_batch_complete=3D8 --iodepth_batch_submit=3D8 --=20 Jens Axboe