From: Wu Fengguang Subject: Re: [patch 3/5] mm: try to distribute dirty pages fairly across zones Date: Mon, 31 Oct 2011 19:33:21 +0800 Message-ID: <20111031113321.GA30890@localhost> References: <1317367044-475-1-git-send-email-jweiner@redhat.com> <1317367044-475-4-git-send-email-jweiner@redhat.com> <20110930142805.GC869@tiehlicka.suse.cz> <20111028201829.GA20607@localhost> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Michal Hocko , Andrew Morton , Mel Gorman , Christoph Hellwig , Dave Chinner , Jan Kara , Rik van Riel , Minchan Kim , Chris Mason , Theodore Ts'o , Andreas Dilger , "Li, Shaohua" , "xfs@oss.sgi.com" , "linux-btrfs@vger.kernel.org" , "linux-ext4@vger.kernel.org" , "linux-mm@kvack.org" , "linux-fsdevel@vger.kernel.org" , "linux-kernel@vger.kernel.org" To: Johannes Weiner Return-path: Content-Disposition: inline In-Reply-To: <20111028201829.GA20607@localhost> Sender: owner-linux-mm@kvack.org List-Id: linux-ext4.vger.kernel.org > //regression > 3) much increased cpu %user and %system for btrfs Sorry I find out that the CPU time regressions for btrfs are caused by some additional trace events enabled on btrfs (for debugging an unrelated btrfs hang bug) which results in 7 times more trace event lines: 2701238 /export/writeback/thresh=1000M/btrfs-1dd-4k-8p-2941M-1000M:10-3.1.0-rc9-ioless-full-nfs-wq5-next-20111014+ 19054054 /export/writeback/thresh=1000M/btrfs-1dd-4k-8p-2941M-1000M:10-3.1.0-rc9-ioless-full-per-zone-dirty-next-20111014+ So no real regressions. Besides, the patchset also performs good on random writes: 3.1.0-rc9-ioless-full-nfs-wq5-next-20111014+ 3.1.0-rc9-ioless-full-per-zone-dirty-next-20111014+ ------------------------ ------------------------ 1.65 -5.1% 1.57 MMAP-RANDWRITE-4K/btrfs-fio_fat_mmap_randwrite_4k-4k-8p-4096M-20:10-X 18.65 -6.4% 17.46 MMAP-RANDWRITE-4K/ext3-fio_fat_mmap_randwrite_4k-4k-8p-4096M-20:10-X 2.09 +1.2% 2.12 MMAP-RANDWRITE-4K/ext4-fio_fat_mmap_randwrite_4k-4k-8p-4096M-20:10-X 2.49 -0.3% 2.48 MMAP-RANDWRITE-4K/xfs-fio_fat_mmap_randwrite_4k-4k-8p-4096M-20:10-X 51.35 +0.0% 51.36 MMAP-RANDWRITE-64K/btrfs-fio_fat_mmap_randwrite_64k-64k-8p-4096M-20:10-X 45.20 +0.5% 45.43 MMAP-RANDWRITE-64K/ext3-fio_fat_mmap_randwrite_64k-64k-8p-4096M-20:10-X 44.77 +0.7% 45.10 MMAP-RANDWRITE-64K/ext4-fio_fat_mmap_randwrite_64k-64k-8p-4096M-20:10-X 45.11 +2.5% 46.23 MMAP-RANDWRITE-64K/xfs-fio_fat_mmap_randwrite_64k-64k-8p-4096M-20:10-X 211.31 +0.2% 211.74 TOTAL write_bw And writes to USB key: 3.1.0-rc9-ioless-full-nfs-wq5-next-20111014+ 3.1.0-rc9-ioless-full-per-zone-dirty-next-20111014+ ------------------------ ------------------------ 5.94 +0.8% 5.99 UKEY-thresh=1G/btrfs-1dd-4k-8p-4096M-1024M:10-X 2.64 -0.8% 2.62 UKEY-thresh=1G/ext3-10dd-4k-8p-4096M-1024M:10-X 5.10 +0.3% 5.12 UKEY-thresh=1G/ext3-1dd-4k-8p-4096M-1024M:10-X 3.26 -0.8% 3.24 UKEY-thresh=1G/ext3-2dd-4k-8p-4096M-1024M:10-X 5.63 -0.5% 5.60 UKEY-thresh=1G/ext4-10dd-4k-8p-4096M-1024M:10-X 6.04 -0.1% 6.04 UKEY-thresh=1G/ext4-1dd-4k-8p-4096M-1024M:10-X 5.90 -0.2% 5.88 UKEY-thresh=1G/ext4-2dd-4k-8p-4096M-1024M:10-X 2.45 +22.6% 3.00 UKEY-thresh=1G/xfs-10dd-4k-8p-4096M-1024M:10-X 6.18 -0.4% 6.16 UKEY-thresh=1G/xfs-1dd-4k-8p-4096M-1024M:10-X 4.81 +0.0% 4.81 UKEY-thresh=1G/xfs-2dd-4k-8p-4096M-1024M:10-X 47.94 +1.1% 48.45 TOTAL write_bw In summary, I see no problem at all in these trivial writeback tests. Tested-by: Wu Fengguang Thanks, Fengguang -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: email@kvack.org