Re: [LKP] [lkp] [f2fs] ec795418c4: fsmark.files_per_sec -36.3% regression

From: Huang\, Ying
Date: Mon Sep 26 2016 - 20:50:17 EST


Jaegeuk Kim <jaegeuk@xxxxxxxxxx> writes:

> On Mon, Sep 26, 2016 at 02:26:06PM +0800, Huang, Ying wrote:
>> Hi, Jaegeuk,
>>
>> "Huang, Ying" <ying.huang@xxxxxxxxx> writes:
>>
>> > Jaegeuk Kim <jaegeuk@xxxxxxxxxx> writes:
>> >
>> >> Hello,
>> >>
>> >> On Sat, Aug 27, 2016 at 10:13:34AM +0800, Fengguang Wu wrote:
>> >>> Hi Jaegeuk,
>> >>>
>> >>> > > >> > - [lkp] [f2fs] b93f771286: aim7.jobs-per-min -81.2% regression
>> >>> > > >> >
>> >>> > > >> > The disk is 4 12G ram disk, and setup RAID0 on them via mdadm. The
>> >>> > > >> > steps for aim7 is,
>> >>> > > >> >
>> >>> > > >> > cat > workfile <<EOF
>> >>> > > >> > FILESIZE: 1M
>> >>> > > >> > POOLSIZE: 10M
>> >>> > > >> > 10 sync_disk_rw
>> >>> > > >> > EOF
>> >>> > > >> >
>> >>> > > >> > (
>> >>> > > >> > echo $HOSTNAME
>> >>> > > >> > echo sync_disk_rw
>> >>> > > >> >
>> >>> > > >> > echo 1
>> >>> > > >> > echo 600
>> >>> > > >> > echo 2
>> >>> > > >> > echo 600
>> >>> > > >> > echo 1
>> >>> > > >> > ) | ./multitask -t &
>> >>> > > >>
>> >>> > > >> Any update on these 2 regressions? Is the information is enough for you
>> >>> > > >> to reproduce?
>> >>> > > >
>> >>> > > > Sorry, I've had no time to dig this due to business travel now.
>> >>> > > > I'll check that when back to US.
>> >>> > >
>> >>> > > Any update?
>> >>> >
>> >>> > Sorry, how can I get multitask binary?
>> >>>
>> >>> It's part of aim7, which can be downloaded here:
>> >>>
>> >>> http://nchc.dl.sourceforge.net/project/aimbench/aim-suite7/Initial%20release/s7110.tar.Z
>> >>
>> >> Thank you for the codes.
>> >>
>> >> I've run this workload on the latest f2fs and compared performance having
>> >> without the reported patch. (1TB nvme SSD, 16 cores, 16GB DRAM)
>> >> Interestingly, I could find slight performance improvement rather than
>> >> regression. :(
>> >> Not sure how to reproduce this.
>> >
>> > I think the difference lies on disk used. The ramdisk is used in the
>> > original test, but it appears that your memory is too small to setup the
>> > RAM disk for test. So it may be impossible for you to reproduce the
>> > test unless you can find more memory :)
>> >
>> > But we can help you to root cause the issue. What additional data do
>> > you want? perf-profile data before and after the patch?
>>
>> Any update to this regression?
>
> Sorry, no. But meanwhile, I've purchased more DRAMs. :)
> Now I'm with 128GB DRAM. I can configure 64GB as pmem.
> Is it worth to try the test again?

I think you are the decision maker for this. You can judge whether the
test is reasonable. And we can adjust our test accordingly.

BTW: For this test, we use brd ram disk and raid to test.

Best Regards,
Huang, Ying