All of lore.kernel.org
 help / color / mirror / Atom feed
* F2FS performance drop than ext4
@ 2013-11-09  2:31 guorke
  2013-11-11  0:35 ` Jaegeuk Kim
  0 siblings, 1 reply; 2+ messages in thread
From: guorke @ 2013-11-09  2:31 UTC (permalink / raw)
  To: linux-f2fs-devel


[-- Attachment #1.1: Type: text/plain, Size: 1216 bytes --]

Dear all,

I test f2fs performace via androbench with lot of f2fs commits.  And i
found that, before commit
bfe35965ecdc6038314d03456b94d9ba451c289d(2013/05/28), the sequential/random
io and sqlite operation performance is better than ext4.  After this
commit, the io performance drop too much,  even test it with the latest
commit.   And test result shows it slower than ext4(~15%), like sequential
read,sqlite insert/update/delete operation,but random write still faster
than ext4(~200%).
F2fs freeze all io operations when do a checkpoint,  but ext4 no this
design,  the checkpoint freeze operation block others metat data update.
in android phone, most application use sqlite , which it caused lot of
fsync/fdatasync. And this sync operation would cause lot of checkpoint
operation , than it caused database perfomance drop. There's some
optimization about writecheckpoint from fs_lock, to semphone. but it is
useless.
any suggestion about it ?

Ps,
1. androbench is  a tool which use to monitor android phone io
performance,  it use fsync/fdatasync to simulate random io and sqlite
database operation.
2. F2fs use default mount option, ext4 use
rw,noatime,noauto_da_alloc,commit=1,data=ordered.


thanks
gourke

[-- Attachment #1.2: Type: text/html, Size: 1340 bytes --]

[-- Attachment #2: Type: text/plain, Size: 435 bytes --]

------------------------------------------------------------------------------
November Webinars for C, C++, Fortran Developers
Accelerate application performance with scalable programming models. Explore
techniques for threading, error checking, porting, and tuning. Get the most 
from the latest Intel processors and coprocessors. See abstracts and register
http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk

[-- Attachment #3: Type: text/plain, Size: 179 bytes --]

_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: F2FS performance drop than ext4
  2013-11-09  2:31 F2FS performance drop than ext4 guorke
@ 2013-11-11  0:35 ` Jaegeuk Kim
  0 siblings, 0 replies; 2+ messages in thread
From: Jaegeuk Kim @ 2013-11-11  0:35 UTC (permalink / raw)
  To: guorke; +Cc: linux-f2fs-devel

Hi,

As I checked the commit, I suspect that you turned on
CONFIG_DEBUG_MUTEXES in your .config.
Could you test without this kind of debugging configs?
Because f2fs uses many mutexes which results in high SW overhead.
Thanks,

2013-11-09 (토), 10:31 +0800, guorke:
> Dear all,
> 
> I test f2fs performace via androbench with lot of f2fs commits.  And i
> found that, before commit
> bfe35965ecdc6038314d03456b94d9ba451c289d(2013/05/28), the
> sequential/random io and sqlite operation performance is better than
> ext4.  After this commit, the io performance drop too much,  even test
> it with the latest commit.   And test result shows it slower than
> ext4(~15%), like sequential read,sqlite insert/update/delete
> operation,but random write still faster than ext4(~200%).    
> F2fs freeze all io operations when do a checkpoint,  but ext4 no this
> design,  the checkpoint freeze operation block others metat data
> update.   in android phone, most application use sqlite , which it
> caused lot of fsync/fdatasync. And this sync operation would cause lot
> of checkpoint operation , than it caused database perfomance drop.
> There's some optimization about writecheckpoint from fs_lock, to
> semphone. but it is useless.
> 
> any suggestion about it ?
> 
> Ps, 
> 1. androbench is  a tool which use to monitor android phone io
> performance,  it use fsync/fdatasync to simulate random io and sqlite
> database operation.
> 2. F2fs use default mount option, ext4 use
> rw,noatime,noauto_da_alloc,commit=1,data=ordered.
> 
> 
> 
> thanks
> 
> gourke
> 
> ------------------------------------------------------------------------------
> November Webinars for C, C++, Fortran Developers
> Accelerate application performance with scalable programming models. Explore
> techniques for threading, error checking, porting, and tuning. Get the most 
> from the latest Intel processors and coprocessors. See abstracts and register
> http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk
> _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

-- 
Jaegeuk Kim
Samsung



------------------------------------------------------------------------------
November Webinars for C, C++, Fortran Developers
Accelerate application performance with scalable programming models. Explore
techniques for threading, error checking, porting, and tuning. Get the most 
from the latest Intel processors and coprocessors. See abstracts and register
http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2013-11-11  0:36 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-11-09  2:31 F2FS performance drop than ext4 guorke
2013-11-11  0:35 ` Jaegeuk Kim

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.