about the rewrite performance of bonnie on ext3cow

[黄荣荣]

hi, I have simply modified ext3cow to let it automatically create a new version of file on every write.
According to the paper of ext3cow on 2005 ACM transaction on storage, the bonnie's rewrite performance of ext3cow is worse than ext3, because the copy-on-write bitmap operations that must be performed when rewriting a buffer.However, when I use bonnie++ to test the performance of the file system, the performance of "rewrite" is better than ext3.
this is ext3's performance:
Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CPmsns710 2016M 10471 98 30663 43 13417 16 13236 99 33837 20 323.4 1

this is ext3cow's performance
Version 1.03 ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CPmsns710 2016M 10476 98 30437 43 24653 44 12066 90 32160 20 +++++ +++

I'm confused about the performance result, can somebody shed some light on this?
_________________________________________________________________
MSN ÖÐÎÄÍø£¬×îÐÂʱÉÐÉú»î×ÊѶ£¬°×ÁìŸÛŒ¯ÃÅ»§¡£
http://cn.msn.com

[Zachary N. J. Peterson]

The ACM paper results were measured on the 2.4 version of ext3cow.
When porting to 2.6 I made a number of performance improvements,
including changing the way COW bitmaps were handled. The COW bitmaps
should now have nearly no effect on performance results.

-Zachary

---
Zachary N. J. Peterson
Assistant Research Scientist
The Johns Hopkins University

***@jhu.edu
http://znjp.com

Post by 黄荣荣
hi, I have simply modified ext3cow to let it automatically create a
new version of file on every write.
According to the paper of ext3cow on 2005 ACM transaction on
storage, the bonnie's rewrite performance of ext3cow is worse than
ext3, because the copy-on-write bitmap operations that must be
performed when rewriting a buffer.
However, when I use bonnie++ to test the performance of the file
system, the performance of "rewrite" is better than ext3.
Version 1.03 ------Sequential Output------ --Sequential
Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --
Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec
%CP /sec %CP
msns710 2016M 10471 98 30663 43 13417 16 13236 99 33837
20 323.4 1
this is ext3cow's performance
Version 1.03 ------Sequential Output------ --Sequential
Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --
Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec
%CP /sec %CP
msns710 2016M 10476 98 30437 43 24653 44 12066 90 32160
20 +++++ +++
I'm confused about the performance result, can somebody shed some light on this?
轻松把Hotmail下载到本地，试试 Windows
Live Mail。 立即尝试！_______________________________________________
ext3cow-devel mailing list
http://hssl.cs.jhu.edu/cgi-bin/mailman/listinfo/ext3cow-devel

[黄荣荣]

hi, thank you for your reply!
But in my bonnie test, the "rewrite" performance of ext3 is 13435 K/sec, and the performance of ext3cow(I simply modify ext3cow to automatically generate a new version on write) is 24798 K/sec.

the "rewrite" performance of ext3 is only 54% of modified ext3cow. I'm confused about the result
paper of ext3cow on 2005 ACM transaction on > > storage, the bonnie's rewrite performance of ext3cow is worse than > > ext3, because the copy-on-write bitmap operations that must be > > performed
when rewriting a buffer.> > However, when I use bonnie++ to test the performance of the file > > system, the performance of "rewrite" is better than ext3.> > this is ext3's performance:> > Version 1.03 ------Sequential Output------ --Sequential > > Input- --Random-> > -Per Chr- --Block-- -Rewrite- -Per Chr- -- > > Block-- --Seeks--> > Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec > > %CP /sec %CP> > msns710 2016M 10471 98 30663 43 13417 16 13236 99 33837 > > 20 323.4 1> >> > this is ext3cow's performance> > Version 1.03 ------Sequential Output------ --Sequential > > Input- --Random-> > -Per Chr- --Block-- -Rewrite- -Per Chr- -- > > Block-- --Seeks--> > Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec > > %CP /sec %CP> > msns710 2016M 10476 98 30437 43 24653 44 12066
90 32160 > > 20 +++++ +++> >> > I'm confused about the performance result, can somebody shed some > > light on this?> >> > ÇáËÉ°ÑHotmailÏÂÔصœ±ŸµØ£¬ÊÔÊÔ Windows > > Live Mail¡£
Á¢ŒŽ³¢ÊÔ£¡_______________________________________________> > ext3cow-devel mailing list> > ext3cow-devel-VAX/***@public.gmane.org> > http://hssl.cs.jhu.edu/cgi-bin/mailman/listinfo/ext3cow-devel>
_________________________________________________________________
Windows Live Photo gallery ÊýÂëÏà»úµÄ³¬Œ¶°é£¬ÇáËɹÜÀíºÍ±àŒ­ÕÕƬ£¬»¹ÄÜÖÆ×÷È«Ÿ°ÃÀÍŒ£¡
http://get.live.cn/product/photo.html

[Zachary N. J. Peterson]

That is curious. I'm not sure how I did that. If you're creating a
new version (by taking a snapshot) is it possible that allocating new
blocks is faster than overwriting old ones? Maybe, especially if the
memory system must page the old blocks back in before overwriting them.

-Zachary

---
Zachary N. J. Peterson
Assistant Research Scientist
The Johns Hopkins University

***@jhu.edu
http://znjp.com

Post by 黄荣荣
hi, thank you for your reply!
But in my bonnie test, the "rewrite" performance of ext3 is 13435 K/
sec, and the performance of ext3cow(I simply modify ext3cow to
automatically generate a new version on write) is 24798 K/sec.
the "rewrite" performance of ext3 is only 54% of modified ext3cow.
I'm confused about the result

Subject: Re: [ext3cow-devel] about the rewrite performance of

bonnie on ext3cow

Date: Mon, 9 Jun 2008 10:53:43 -0400
The ACM paper results were measured on the 2.4 version of ext3cow.
When porting to 2.6 I made a number of performance improvements,
including changing the way COW bitmaps were handled. The COW bitmaps
should now have nearly no effect on performance results.
-Zachary
---
Zachary N. J. Peterson
Assistant Research Scientist
The Johns Hopkins University
http://znjp.com

Post by 黄荣荣
hi, I have simply modified ext3cow to let it automatically

create a

Post by 黄荣荣
new version of file on every write.
According to the paper of ext3cow on 2005 ACM transaction on
storage, the bonnie's rewrite performance of ext3cow is worse than
ext3, because the copy-on-write bitmap operations that must be
performed when rewriting a buffer.
However, when I use bonnie++ to test the performance of the file
system, the performance of "rewrite" is better than ext3.
Version 1.03 ------Sequential Output------ --Sequential
Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per Chr- --
Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec
%CP /sec %CP
msns710 2016M 10471 98 30663 43 13417 16 13236 99 33837
20 323.4 1
this is ext3cow's performance
Version 1.03 ------Sequential Output------ --Sequential
Input- --Random-
-Per Chr- --Block-- -Rewrite- -Per C hr- --
Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec
%CP /sec %CP
msns710 2016M 10476 98 30437 43 24653 44 12066 90 32160
20 +++++ +++
I'm confused about the performance result, can somebody shed some light on this?
轻松把Hotmail下载到本地，试试 Windows
Live Mail。 立即尝试！

_______________________________________________

Post by 黄荣荣
ext3cow-devel mailing list
http://hssl.cs.jhu.edu/cgi-bin/mailman/listinfo/ext3cow-devel

使用新一代 Windows Live Messenger 轻松交流和共享！ 立即体验！