[nSLUG] 2.6 kernel performance

Donald Teed donald.teed at gmail.com
Wed Jun 22 03:34:50 ADT 2005


I've run 2.6.8 kernel with a few variations. Tried preempt option as well - 
not a great deal of difference.

I've also googled, and discovered there are many people reporting
IDE performs slower on 2.6 than on 2.4, regardless of what
scheduler you set up, or what hdparm args you throw at it.

Here are bonnie results in my case, using a raid 5 partition and a normal
partition at the end for comparison:

2.6.8:
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
artistic 1G 8105 95 32693 54 12825 23 8821 92 24762 24 260.4 3
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 496 99 +++++ +++ 23610 95 522 98 +++++ +++ 1927 97
artistic,1G,8105,95,32693,54,12825,23,8821,92,24762,24,260.4
,3,16,496,99,+++++,+++,23610,95,522,98,+++++,+++,1927,97


2.6.8 with preempt in kernel:
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
artistic 1G 8288 98 39729 77 14980 29 8861 93 29514 30 321.3 3
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 496 99 +++++ +++ 23249 99 513 99 +++++ +++ 1954 97
artistic,1G,8288,98,39729,77,14980,29,8861,93,29514,30,321.3
,3,16,496,99,+++++,+++,23249,99,513,99,+++++,+++,1954,97


2.4.31:
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
artistic 1G 5803 100 35207 79 15762 46 7227 98 64142 51 331.6 3
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 490 98 +++++ +++ 20328 100 539 98 +++++ +++ 2195 97
artistic,1G,5803,100,35207,79,15762,46,7227,98,64142,51,331.6
,3,16,490,98,+++++,+++,20328,100,539,98,+++++,+++,2195,97

2.4.31 on non-RAID partitoin:
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
artistic 1G 6273 99 46759 82 18065 26 7184 96 48394 28 191.3 1
------Sequential Create------ --------Random Create--------
-Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 534 99 +++++ +++ 21957 100 546 99 +++++ +++ 2200 99
artistic,1G,6273,99,46759,82,18065,26,7184,96,48394,28,191.3
,1,16,534,99,+++++,+++,21957,100,546,99,+++++,+++,2200,99

The most significant performance variant I spot there are sequential input 
blocks at 64142 K/sec -
about double of that for 2.6 kernel. There are many numbers which are 
showing
slightly higher for 2.6 with preempt compared to the 2.4 kernel.

I think this benchmark leaves me with little information to go by. If I 
don't know the
internals, for example, of how an IMAP server opens a mailbox, how can I 
compare?


!DSPAM:42b9068935551120920689!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://nslug.ns.ca/mailman/private/nslug/attachments/20050622/1abbb791/attachment-0002.html>


More information about the nSLUG mailing list