I decided to try btrfs on F13 (2.6.33.6-147.2.4.fc13.x86_64 kernel) with the
following fs_mark command and a 1.5 TB Seagate S-ATA disk:
# fs_mark -s 0 -S 0 -D 1000 -n 1000000 -L 1000 -d /test/ -l btrfs_log.txt
btrfs starts off at a fantastic rate - roughly 3-4 times the speed of ext4:
FSUse% Count Size Files/sec App Overhead
0 1000000 0 20815.9 6257344
0 2000000 0 17531.2 6310061
0 3000000 0 10656.4 6473972
0 4000000 0 17933.6 6291377
0 5000000 0 6687.9 6708750
0 6000000 0 6494.3 6792701
0 7000000 0 18211.1 6266870
0 8000000 0 18518.4 6231522
Then it chugs along, but hits relatively long periods of slowness:
20 231000000 0 16991.8 6216216
20 232000000 0 17524.1 6290540
20 233000000 0 17088.5 6456494
20 234000000 0 1259.1 10185066
20 235000000 0 1419.0 8144068
20 236000000 0 1670.9 12154381
20 237000000 0 1863.6 8582337
20 238000000 0 929.1 9714860
21 239000000 0 4100.7 7385278
21 240000000 0 3486.6 8773568
21 241000000 0 2224.7 7320514
21 242000000 0 3761.8 7166617
21 243000000 0 3197.0 6918920
21 244000000 0 1293.1 8486926
21 245000000 0 3189.7 7396151
21 246000000 0 3164.7 7103912
21 247000000 0 4028.8 6770544
21 248000000 0 3698.7 7133084
21 249000000 0 2247.9 7421000
22 250000000 0 1579.3 9833236
I will fill this and get some plots, etc to do relative performance & then
wanted to try the current RC kernel just to refresh.
Ric
(Whole log file appended)
On Mon, Aug 16, 2010 at 08:29:24AM -0400, Ric Wheeler wrote:> > I decided to try btrfs on F13 (2.6.33.6-147.2.4.fc13.x86_64 kernel) > with the following fs_mark command and a 1.5 TB Seagate S-ATA disk: > > # fs_mark -s 0 -S 0 -D 1000 -n 1000000 -L 1000 -d /test/ -l btrfs_log.txt > > btrfs starts off at a fantastic rate - roughly 3-4 times the speed of ext4: >[ log ]> > I will fill this and get some plots, etc to do relative performance > & then wanted to try the current RC kernel just to refresh.Thanks Ric, I think this is related to the perf problem delalloc flushing performance bug. I''ll nail it down shortly. -chris -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 08/16/2010 08:37 AM, Chris Mason wrote:> On Mon, Aug 16, 2010 at 08:29:24AM -0400, Ric Wheeler wrote: >> I decided to try btrfs on F13 (2.6.33.6-147.2.4.fc13.x86_64 kernel) >> with the following fs_mark command and a 1.5 TB Seagate S-ATA disk: >> >> # fs_mark -s 0 -S 0 -D 1000 -n 1000000 -L 1000 -d /test/ -l btrfs_log.txt >> >> btrfs starts off at a fantastic rate - roughly 3-4 times the speed of ext4: >> > [ log ] > >> I will fill this and get some plots, etc to do relative performance >> & then wanted to try the current RC kernel just to refresh. > Thanks Ric, I think this is related to the perf problem delalloc > flushing performance bug. I''ll nail it down shortly. > > -chrisGreat, I will be happy to test patches when you have them. No particular hurry on my end. I also hope to get our new arrays online and should be able to steal a couple of days run there with luck, Ric -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Aug 16, 2010 at 08:29:24AM -0400, Ric Wheeler wrote:> > I decided to try btrfs on F13 (2.6.33.6-147.2.4.fc13.x86_64 kernel) > with the following fs_mark command and a 1.5 TB Seagate S-ATA disk: > > # fs_mark -s 0 -S 0 -D 1000 -n 1000000 -L 1000 -d /test/ -l btrfs_log.txt > > btrfs starts off at a fantastic rate - roughly 3-4 times the speed of ext4:The results are interesting, both ext4 and btrfs seem to have a little more than a million inodes in slab cache on my box. The ext4 inode is a little fatter so it''s about 2.2GB instead of 1.7GB. I''ll let btrfs run overnight. ext4 isn''t doing many reads at all but is instead stuck in the log, but there''s very little IO. I''m not actually sure what it is doing. Btrfs is doing a healthy amount of writes and batches of reads. -chris -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html