Hi Allan,

FSS won't work correctly if you have a mix of processes/threads from 
different scheduling classes (TS/IA/FX) in the same processor set, since 
those classes use the same priority range (0-59). Might that be 
happening here?

Thanks,
-Eric

Allan wrote:
> Hi,
>
> I am setting up the FSS within 3 projects, my understanding is that if I have 
> project a=1 share b=2 and c=3 , then the cpu amount if there is contention 
> should be c=50% b=33% and c=17%. Server is a E2900 with 8 cpus.
>
> When I run a test using the following code
>
> foreach $i (1..30) {
>                 $pid = fork();
>                 last if $pid == 0;
>                 print "created PID $pid\n";
>         }
>
>         while (1) {
>                 $x++;
>         }
>
> Where 1..3 = project a , 1..60 for b and 1..120 for c , I would assume that I 
> would hit the limits for each projects shares and see from prstat -J the 
> correct limits being attained. Can you adivse wither I am looking at this 
> incorrectly? or are the tests I am using flawed?
>
> prstat 
>  10155 root     4048K 2480K run     41    4   0:01:45 1.1% perl/1
>  10166 root     4040K 1296K run     43    4   0:01:42 1.1% perl/1
>  10175 root     4040K 1296K run     43    4   0:01:42 1.1% perl/1
>  10171 root     4040K 1296K run     44    4   0:01:41 1.1% perl/1
>  10172 root     4040K 1296K run     44    4   0:01:43 1.1% perl/1
>  10176 root     4040K 1296K run     44    4   0:01:43 1.1% perl/1
>  10178 root     4040K 1296K run     43    4   0:01:44 1.1% perl/1
>  10156 root     4040K 1440K run     43    4   0:01:44 1.1% perl/1
>  10168 root     4040K 1296K run     43    4   0:01:41 1.1% perl/1
>  10169 root     4040K 1296K cpu9    46    4   0:01:44 1.0% perl/1
>  10167 root     4040K 1296K run     43    4   0:01:41 1.0% perl/1
>  10184 root     4040K 1296K run     43    4   0:01:43 1.0% perl/1
>  10162 root     4040K 1296K run     44    4   0:01:42 1.0% perl/1
>  10183 root     4040K 1296K run     44    4   0:01:40 1.0% perl/1
>  10179 root     4040K 1296K run     43    4   0:01:41 1.0% perl/1
>  10157 root     4040K 1296K run     46    4   0:01:44 1.0% perl/1
>  10163 root     4040K 1296K run     44    4   0:01:45 1.0% perl/1
>  10159 root     4040K 1296K run     45    4   0:01:43 1.0% perl/1
>  10165 root     4040K 1296K run     44    4   0:01:41 1.0% perl/1
>  10161 root     4040K 1296K run     45    4   0:01:44 1.0% perl/1
>  10170 root     4040K 1296K cpu513  47    4   0:01:40 1.0% perl/1
>  10181 root     4040K 1296K cpu1    45    4   0:01:41 1.0% perl/1
>  10180 root     4040K 1296K cpu523  47    4   0:01:43 1.0% perl/1
>  10164 root     4040K 1296K run     46    4   0:01:42 1.0% perl/1
>  10160 root     4040K 1296K cpu514  46    4   0:01:41 1.0% perl/1
>  10158 root     4040K 1296K cpu515  47    4   0:01:41 1.0% perl/1
>  10185 root     4040K 1296K run     45    4   0:01:41 1.0% perl/1
>  10174 root     4040K 1296K run     46    4   0:01:41 1.0% perl/1
>  10182 root     4040K 1296K run     45    4   0:01:43 1.0% perl/1
>  10173 root     4040K 1296K run     45    4   0:01:42 1.0% perl/1
> PROJID    NPROC  SIZE   RSS MEMORY      TIME  CPU PROJECT
>    110      121  477M  154M   0.0%   0:47:55  34% cpu.unallocated
>    100       61  241M   79M   0.0%   0:48:01  32% cpu.spike
>    120       31  122M   41M   0.0%   0:52:51  32% allan
>      1        9   45M   26M   0.0%   0:00:17 0.0% user.root
>      0       43  478M  297M   0.4%   0:27:51 0.0% system
>
>
>
> Thanks
>  
>  
> This message posted from opensolaris.org
> _______________________________________________
> perf-discuss mailing list
> perf-discuss@opensolaris.org
>   

_______________________________________________
perf-discuss mailing list
perf-discuss@opensolaris.org

Reply via email to