lkml.org 
[lkml]   [2012]   [Feb]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: getdents - ext4 vs btrfs performance
From
2012/2/29 Jacek Luczak <difrost.kernel@gmail.com>:
> 2012/2/29 Jacek Luczak <difrost.kernel@gmail.com>:
>> Hi Chris,
>>
>> the last one was borked :) Please check this one.
>>
>> -jacek
>>
>> 2012/2/29 Jacek Luczak <difrost.kernel@gmail.com>:
>>> Hi All,
>>>
>>> /*Sorry for sending incomplete email, hit wrong button :) I guess I
>>> can't use Gmail */
>>>
>>> Long story short: We've found that operations on a directory structure
>>> holding many dirs takes ages on ext4.
>>>
>>> The Question: Why there's that huge difference in ext4 and btrfs? See
>>> below test results for real values.
>>>
>>> Background: I had to backup a Jenkins directory holding workspace for
>>> few projects which were co from svn (implies lot of extra .svn dirs).
>>> The copy takes lot of time (at least more than I've expected) and
>>> process was mostly in D (disk sleep). I've dig more and done some
>>> extra test to see if this is not a regression on block/fs site. To
>>> isolate the issue I've also performed same tests on btrfs.
>>>
>>> Test environment configuration:
>>> 1) HW: HP ProLiant BL460 G6, 48 GB of memory, 2x 6 core Intel X5670 HT
>>> enabled, Smart Array P410i, RAID 1 on top of 2x 10K RPM SAS HDDs.
>>> 2) Kernels: All tests were done on following kernels:
>>>  - 2.6.39.4-3 -- the build ID (3) is used here for internal tacking of
>>> config changes mostly. In -3 we've introduced ,,fix readahead pipeline
>>> break caused by block plug'' patch. Otherwise it's pure 2.6.39.4.
>>>  - 3.2.7 -- latest kernel at the time of testing (3.2.8 has been
>>> release recently).
>>> 3) A subject of tests, directory holding:
>>>  - 54GB of data (measured on ext4)
>>>  - 1978149 files
>>>  - 844008 directories
>>> 4) Mount options:
>>>  - ext4 -- errors=remount-ro,noatime,
>>> data=writeback
>>>  - btrfs -- noatime,nodatacow and for later investigation on
>>> copression effect: noatime,nodatacow,compress=lzo
>>>
>>> In all tests I've been measuring time of execution. Following tests
>>> were performed:
>>> - find . -type d
>>> - find . -type f
>>> - cp -a
>>> - rm -rf
>>>
>>> Ext4 results:
>>> | Type     | 2.6.39.4-3   | 3.2.7
>>> | Dir cnt  | 17m 40sec  | 11m 20sec
>>> | File cnt |  17m 36sec | 11m 22sec
>>> | Copy    | 1h 28m        | 1h 27m
>>> | Remove| 3m 43sec    | 3m 38sec
>>>
>>> Btrfs results (without lzo comression):
>>> | Type     | 2.6.39.4-3   | 3.2.7
>>> | Dir cnt  | 2m 22sec  | 2m 21sec
>>> | File cnt |  2m 26sec | 2m 23sec
>>> | Copy    | 36m 22sec | 39m 35sec
>>> | Remove| 7m 51sec   | 10m 43sec
>>>
>>> From above one can see that copy takes close to 1h less on btrfs. I've
>>> done strace counting times of calls, results are as follows (from
>>> 3.2.7):
>>> 1) Ext4 (only to elements):
>>> % time     seconds  usecs/call     calls    errors syscall
>>> ------ ----------- ----------- --------- --------- ----------------
>>>  57.01   13.257850           1  15082163           read
>>>  23.40    5.440353           3   1687702           getdents
>>>  6.15    1.430559           0   3672418           lstat
>>>  3.80    0.883767           0  13106961           write
>>>  2.32    0.539959           0   4794099           open
>>>  1.69    0.393589           0    843695           mkdir
>>>  1.28    0.296700           0   5637802           setxattr
>>>  0.80    0.186539           0   7325195           stat
>>>
>>> 2) Btrfs:
>>> % time     seconds  usecs/call     calls    errors syscall
>>> ------ ----------- ----------- --------- --------- ----------------
>>> 53.38    9.486210           1  15179751           read
>>> 11.38    2.021662           1   1688328           getdents
>>>  10.64    1.890234           0   4800317           open
>>>  6.83    1.213723           0  13201590           write
>>>  4.85    0.862731           0   5644314           setxattr
>>>  3.50    0.621194           1    844008           mkdir
>>>  2.75    0.489059           0   3675992         1 lstat
>>>  1.71    0.303544           0   5644314           llistxattr
>>>  1.50    0.265943           0   1978149           utimes
>>>  1.02    0.180585           0   5644314    844008 getxattr
>>>
>>> On btrfs getdents takes much less time which prove the bottleneck in
>>> copy time on ext4 is this syscall. In 2.6.39.4 it shows even less time
>>> for getdents:
>>> % time     seconds  usecs/call     calls    errors syscall
>>> ------ ----------- ----------- --------- --------- ----------------
>>>  50.77   10.978816           1  15033132           read
>>>  14.46    3.125996           1   4733589           open
>>>  7.15    1.546311           0   5566988           setxattr
>>>  5.89    1.273845           0   3626505           lstat
>>>  5.81    1.255858           1   1667050           getdents
>>>  5.66    1.224403           0  13083022           write
>>>  3.40    0.735114           1    833371           mkdir
>>>  1.96    0.424881           0   5566988           llistxattr
>>>
>>>
>>> Why so huge difference in the getdents timings?
>>>
>>> -Jacek
>
> I will try to answer the question from the broken email I've sent.
>
> @Lukas, it was always a fresh FS on top of LVM logical volume. I've
> been cleaning cache/remounting to sync all data before (re)doing
> tests.
>
> -Jacek
>
> BTW: Sorry for the email mixture. I just can't get this gmail thing to
> work (why forcing top posting:/). Please use this thread.

More from the observations:
1) 10s dump of the process state during copy shows:
- Ext4: 526 probes done, 34 hits R state, 492 hits D state
- Btrfs (2.6.39.4): 218, 83, 135
- Btrfs (3.2.7): 238, 62, 174, 2 hit sleeping
2) dd write/read of 55GB file to/from volume:
- Ext4: write 127MB/s, read 107MB/s
- Btrfs: 110MB/s, read 176MB/s

-Jacek
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2012-02-29 15:23    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site