lkml.org 
[lkml]   [2006]   [Jun]   [6]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateTue, 6 Jun 2006 10:23:29 +0000 (GMT)
FromHolger Kiehl <>
SubjectQuestion regarding ext3 extents+mballoc+delalloc
Hello

Looking at ways to increase write performance on my system using ext3
Andreas Dilger pointed me to delalloc+mballoc+extent patches. Downloaded
those from ftp://ftp.clusterfs.com/pub/people/alex/2.6.16.8 and run some
benchmark, here some results using bonnie++:

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
2.6.16.19       16G 59223  91 264155  45 111459  36 57313  99 317944  63  1478   7
                     58814  92 276803  47 110418  36 57105  99 317534  65  1525   5
                     58299  92 274523  48 110290  36 56723  99 318839  65  1502   4
And here the results when mounting without extents,mballoc,delalloc option:


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
2.6.16.19       16G 38621  98 194816  94 87776  49 37921  92 239128  54  1402   5
                     47000  99 194276  94 89232  49 38628  92 240539  55  1399   5
                     45873  98 178195  90 89726  50 38482  92 240490  55  1381   4
So using delalloc+mballoc+extent gives an approx. 30% increase in performance.

Using my own benchmark afdbench where many process copy thousands of files
around, the results are as follows:

With extents,mballoc,delalloc

      5488.36 files per second 14.19 MB/s

Without extents,mballoc,delalloc

      4829.84 files per second 12.72 MB/s

The performance increase is approx. 12%.

So the question is, why are these patches not included into the kernel?
I did some very extensive testing for several days and could not discover
any disadvantage using those patches. I must add that I did not test
crashes to see if data is lost. Are there any disadvantages using these
patches?

Please CC me since I am not subscribed to the list.

Thanks,
Holger
-- 

-
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: 2006-06-06 12:28    [from the cache]
©2003-2008