From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jamie Lokier Subject: Re: [PATCH 8/8] vm: Add an tuning knob for vm.max_writeback_pages Date: Wed, 2 Sep 2009 00:56:45 +0100 Message-ID: <20090901235645.GD1321@shareable.org> References: <1251803946-9243-1-git-send-email-jens.axboe@oracle.com> <1251803946-9243-9-git-send-email-jens.axboe@oracle.com> <1251830335.8502.17.camel@laptop> <20090901184455.GA27294@infradead.org> <20090901235218.GC1321@shareable.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Peter Zijlstra , Jens Axboe , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, chris.mason@oracle.com, david@fromorbit.com, tytso@mit.edu, akpm@linux-foundation.org, jack@suse.cz To: Christoph Hellwig Return-path: Content-Disposition: inline In-Reply-To: <20090901235218.GC1321@shareable.org> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org Jamie Lokier wrote: > Christoph Hellwig wrote: > > On Tue, Sep 01, 2009 at 08:38:55PM +0200, Peter Zijlstra wrote: > > > Do we really need a tunable for this? > > > > It will make increasing it in the field a lot easier. And having deal > > with really large systems I have the fear that there are I/O topologies > > outhere for which every "reasonable" value is too low. > > > > > I guess we need a limit to avoid it writing out everything, but can't we > > > have something automagic? > > > > Some automatic adjustment would be nice. But finding the right auto > > tuning will be an interesting exercise. > > I have embedded systems with 32MB RAM and no MMU, where I deliberately > make the equivalent of max_writeback_pages *smaller* to limit the > number of dirty pages causing fragmentation and preventing allocation > of high-order pages... Write performance is less important than being > able to allocate contiguous memory for reads. > > They are still using 2.4 kernels, but the principle still applies. > maybe even more on 2.6 which is more prone to fragmentation on small > no-MMU devices. Sorry, I must get more sleep. I confused max_writeback_pages with the limit on dirty pages in the system, which is completely different. So please ignore my previous mail. *little*, -- Jamie