Re: IO CPU affinity test results

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Alan D. Brunelle
Date: Friday, March 14, 2008 - 8:55 am

4-way IA64, better, but not good after 30 runs (in single user mode):

Part  RQ   MIN     AVG     MAX      Dev
----- --  ------  ------  ------  ------
 mkfs 0  18.893  19.429  20.129   0.351
 mkfs 1  18.912  19.347  19.934   0.258

untar 0  17.007  17.538  18.829   0.416
untar 1  16.980  17.329  18.369   0.284

 make 0  22.050  23.000  23.546   0.400
 make 1  22.416  23.141  23.637   0.349

 comb 0  58.138  59.967  61.923   0.793
 comb 1  58.428  59.817  60.750   0.608

 psys 0   7.69%   8.77%   9.93%   0.443
 psys 1   8.47%  14.20%  16.33%   2.566

Better run time w/ rq=1 (only 0.25%) but at a huge cost in %sys. Will do some profiling to see what's up.

Alan
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
IO CPU affinity test results, Alan D. Brunelle, (Fri Mar 14, 5:14 am)
Re: IO CPU affinity test results, Alan D. Brunelle, (Fri Mar 14, 5:23 am)
Re: IO CPU affinity test results, Jens Axboe, (Fri Mar 14, 5:36 am)
Re: IO CPU affinity test results, Jens Axboe, (Fri Mar 14, 5:37 am)
Re: IO CPU affinity test results, Jens Axboe, (Fri Mar 14, 6:05 am)
Re: IO CPU affinity test results, Alan D. Brunelle, (Fri Mar 14, 6:41 am)
Re: IO CPU affinity test results, Alan D. Brunelle, (Fri Mar 14, 8:55 am)