From: Chris Mason Subject: Re: EXT4 nodelalloc => back to stone age. Date: Mon, 1 Apr 2013 11:57:05 -0400 Message-ID: <20130401155705.16754.31446@localhost.localdomain> References: <87d2uese6t.fsf@openvz.org> <5159A55B.1090302@redhat.com> <20130401154541.16754.22860@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 8BIT Cc: ext4 development , "linux-fsdevel@vger.kernel.org" , "axboe@kernel.dk" , Jan Kara To: Chris Mason , Eric Sandeen , Dmitry Monakhov Return-path: Received: from dkim1.fusionio.com ([66.114.96.53]:48741 "EHLO dkim1.fusionio.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758527Ab3DAP5K convert rfc822-to-8bit (ORCPT ); Mon, 1 Apr 2013 11:57:10 -0400 Received: from mx1.fusionio.com (unknown [10.101.1.160]) by dkim1.fusionio.com (Postfix) with ESMTP id ADB037C0682 for ; Mon, 1 Apr 2013 09:57:09 -0600 (MDT) In-Reply-To: <20130401154541.16754.22860@localhost.localdomain> Sender: linux-ext4-owner@vger.kernel.org List-ID: Quoting Chris Mason (2013-04-01 11:45:41) > Quoting Eric Sandeen (2013-04-01 11:18:51) > > On 4/1/13 6:06 AM, Dmitry Monakhov wrote: > > > > > 1)Do we really have to use WRITE_SYNC in case of WB_SYNC_ALL ? > > Yes? The stuff we wait on should be WRITE_SYNC. > > > > > ... > > > > > 2) Why don't we have writepages for non delalloc case ? > > > > ... > > > > I'd add: > > > > 3) Why do we have a "nodelalloc" mount option at all? > > > > but then I thought: > > > > Is it also this bad when using the ext4 driver to run an ext3 fs? > > Quick comparison on a single iodrive: On the theory that writepages is the problem try echo 1 > /sys/block/xxx/queue/rotational. With request merging on here in nodelalloc mode: dd if=/dev/zero of=foo bs=1M count=1024 conv=fsync,notrunc 1073741824 bytes (1.1 GB) copied, 2.53741 s, 423 MB/s dd if=/dev/zero of=foo bs=1M count=1024 conv=fsync,notrunc 1073741824 bytes (1.1 GB) copied, 1.37795 s, 779 MB/s -chris