From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sonny Rao Subject: Re: Re: [RFC] ext3 writepages for writeback mode Date: Mon, 14 Feb 2005 15:02:56 -0500 Message-ID: <20050214200256.GA3690@kevlar.burdell.org> References: <1108085493.20053.1191.camel@dyn318077bld.beaverton.ibm.com> <20050210175327.7b4a508b.akpm@osdl.org> <1108164542.20053.1232.camel@dyn318077bld.beaverton.ibm.com> <20050211155858.354520bb.akpm@osdl.org> <1108166986.20053.1235.camel@dyn318077bld.beaverton.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Andrew Morton , ext2-devel , linux-fsdevel@vger.kernel.org, sct@redhat.com To: Badari Pulavarty Content-Disposition: inline In-Reply-To: <1108166986.20053.1235.camel@dyn318077bld.beaverton.ibm.com> Sender: ext2-devel-admin@lists.sourceforge.net Errors-To: ext2-devel-admin@lists.sourceforge.net List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , List-Archive: List-Id: linux-fsdevel.vger.kernel.org On Fri, Feb 11, 2005 at 04:09:46PM -0800, Badari Pulavarty wrote: > On Fri, 2005-02-11 at 15:58, Andrew Morton wrote: > > Badari Pulavarty wrote: > > > > > > Due to lack of interesting suggestions to solve > > > mpage_writepages() -> ext3_writeback_writepage() problem, > > > I fixed it in the dumbest possible way. > > > > I've actually forgotten what the problem was. It was 100 patches ago :( > > The problem was > ext3_writeback_writepages() -> mpage_writepages() could > call back ext3_writeback_writepage() in the "confused" case. > ext3_writeback_writepage() could end up doing nothing, since the > we already have a journal handle. > > I added a writepage_helper to handle this case. > > > > > > Let me know, what you think. > > > > > > If it works, let's get some benchmark numbers so we can decide whether it > > justifies more development? > > > > Yep. I will get some numbers to see .. > I'm helping Badari collecting data on this... My setup is a P4 2.0Ghz booted with 1GB of RAM and 1 cpu attached via Fiber to a seven disk raid0 array with write-caching turned off (write-cacheing can skew numbers significantly if you aren't careful to let the cache drain between runs, etc.) The test is a single-threaded sequential overwrite of a 20GB data set divided into 512MB files which are selected randomly and overwritten. All of these numbers represent the average of three five-minute runs. All ext3 tests are in writeback mode FS Throughput Cpu Utilizaiton -- ---------- --------------- Ext3 78 MB/sec 75.9 % Ext3 + wpages 85 MB/sec 74.7 % Just for comparison: Ext2 88.5 MB/sec 74.2 % Ext2 + nobh 89.6 MB/sec 71.7 % JFS 94.8 MB/sec 85.6 % XFS 100 MB/sec 95.5 % So, Badari's writepages patch improves performance on this particular setup by almost 10 % I can rerun with more processors/ram, or different disk configurations if anyone is interested. Sonny ------------------------------------------------------- SF email is sponsored by - The IT Product Guide Read honest & candid reviews on hundreds of IT Products from real users. Discover which products truly live up to the hype. Start reading now. http://ads.osdn.com/?ad_id=6595&alloc_id=14396&op=click