On Tue, 2015-03-31 at 15:59 +0100, Filipe Manana wrote:
[snip]
> From: Filipe Manana <[email protected]>
> Date: Tue, 31 Mar 2015 14:16:52 +0100
> Subject: [PATCH] Btrfs: avoid syncing log in the fast fsync path when not
> necessary
>
> Commit 3a8b36f37806 ("Btrfs: fix data loss in the fast fsync path") added
> a performance regression for that causes an unnecessary sync of the log
> trees (fs/subvol and root log trees) when 2 consecutive fsyncs are done
> against a file, without no writes or any metadata updates to the inode in
> between them and if a transaction is committed before the second fsync is
> called.
>
> Huang Ying reported this to lkml after a test sysbench test that measured
> a -62% decrease of file io requests for that tests' workload.
>
> The test is:
>
> echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
> echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
> echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
> echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
> mkfs -t btrfs /dev/sda2
> mount -t btrfs /dev/sda2 /fs/sda2
> cd /fs/sda2
> for ((i = 0; i < 1024; i++)); do fallocate -l 67108864 testfile.$i; done
> sysbench --test=fileio --max-requests=0 --num-threads=4 --max-time=600 \
> --file-test-mode=rndwr --file-total-size=68719476736 --file-io-mode=sync \
> --file-num=1024 run
>
> A test on kvm guest, running a debug kernel gave me the following results:
>
> Without 3a8b36f378060d: 16.01 reqs/sec
> With 3a8b36f378060d: 3.39 reqs/sec
> With 3a8b36f378060d and this patch: 16.04 reqs/sec
>
> Reported-by: Huang Ying <[email protected]>
I have tested your patch, the regression restored in our test. Thanks!
Tested-by: Huang, Ying <[email protected]>
Best Regards,
Huang, Ying
[snip]
On 04/01/2015 05:59 AM, Huang Ying wrote:
> On Tue, 2015-03-31 at 15:59 +0100, Filipe Manana wrote:
>
> [snip]
>
>> From: Filipe Manana <[email protected]>
>> Date: Tue, 31 Mar 2015 14:16:52 +0100
>> Subject: [PATCH] Btrfs: avoid syncing log in the fast fsync path when not
>> necessary
>>
>> Commit 3a8b36f37806 ("Btrfs: fix data loss in the fast fsync path") added
>> a performance regression for that causes an unnecessary sync of the log
>> trees (fs/subvol and root log trees) when 2 consecutive fsyncs are done
>> against a file, without no writes or any metadata updates to the inode in
>> between them and if a transaction is committed before the second fsync is
>> called.
>>
>> Huang Ying reported this to lkml after a test sysbench test that measured
>> a -62% decrease of file io requests for that tests' workload.
>>
>> The test is:
>>
>> echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
>> echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
>> echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor
>> echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor
>> mkfs -t btrfs /dev/sda2
>> mount -t btrfs /dev/sda2 /fs/sda2
>> cd /fs/sda2
>> for ((i = 0; i < 1024; i++)); do fallocate -l 67108864 testfile.$i; done
>> sysbench --test=fileio --max-requests=0 --num-threads=4 --max-time=600 \
>> --file-test-mode=rndwr --file-total-size=68719476736 --file-io-mode=sync \
>> --file-num=1024 run
>>
>> A test on kvm guest, running a debug kernel gave me the following results:
>>
>> Without 3a8b36f378060d: 16.01 reqs/sec
>> With 3a8b36f378060d: 3.39 reqs/sec
>> With 3a8b36f378060d and this patch: 16.04 reqs/sec
>>
>> Reported-by: Huang Ying <[email protected]>
>
> I have tested your patch, the regression restored in our test. Thanks!
>
> Tested-by: Huang, Ying <[email protected]>
Thank you very much for testing it and the report.
I'll now send the exact same patch, with your Tested-by tag, to the
btrfs mailing list, so that Chris can pick it from patchwork and can
more easily be noticed by the btrfs community (for review, test, etc).
regards,
Filipe Manana
>
> Best Regards,
> Huang, Ying
>
> [snip]
>
>
>