From: Theodore Tso Subject: Re: [PATCH, RFC] vm: Add an tuning knob for vm.max_writeback_pages Date: Tue, 1 Sep 2009 16:30:22 -0400 Message-ID: <20090901203022.GD6996@mit.edu> References: <1251600858-21294-1-git-send-email-tytso@mit.edu> <20090830165229.GA5189@infradead.org> <20090830181731.GA20822@mit.edu> <20090901180052.GA7885@think> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii To: Chris Mason , Christoph Hellwig , linux-mm@kvack.org, Ext4 Developers List , linux-fsdevel@vger.kernel.org, jens Return-path: Received: from THUNK.ORG ([69.25.196.29]:59114 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752055AbZIAUaW (ORCPT ); Tue, 1 Sep 2009 16:30:22 -0400 Content-Disposition: inline In-Reply-To: <20090901180052.GA7885@think> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Tue, Sep 01, 2009 at 02:00:52PM -0400, Chris Mason wrote: > > I haven't yet tried this without the max_writeback_pages patch, but the > graphs clearly show a speed improvement, and that the mainline code is > smearing writes across the drive while Jens' work is writing > sequentially. FYI, you don't need to revert the max_writebacks_pages patch; the whole point of making it a tunable was to make it easier to run benchmarks. If you want to get the effects of the original setting of MAX_WRITEBACK_PAGES before the patch, just run as root: sysctl vm.max_writeback_pages=1024 - Ted