[gridengine users] strange cpu time and wallclock

Mike Serkov serkov.m at gmail.com
Thu Dec 7 17:07:25 UTC 2017


Jerome,

It just means that your job used more than one CPU ( probably multithreaded task, or something is running In parallel) SGE doesn't limit your job to use more CPUs than slots requested ( in general - the slot is abstract and just a place on the server for the job, not  physical CPU. ).

Best regards,
Mikhail Serkov

Sent from my iPhone

> On Dec 7, 2017, at 11:50 AM, Jerome <jerome at ibt.unam.mx> wrote:
> 
> Dear all.
> 
> I've a bit confusiong about some accounting data get back from a single job:
> 
> qname        all.q
> hostname     compute-0-4.local
> 
> ../..
> 
> slots        1
> failed       0
> exit_status  0
> ru_wallclock 1761
> ru_utime     2066.817
> ru_stime     104.779
> ru_maxrss    56872
> ru_ixrss     0
> ru_ismrss    0
> ru_idrss     0
> ru_isrss     0
> ru_minflt    227175
> ru_majflt    0
> ru_nswap     0
> ru_inblock   0
> ru_oublock   51524248
> ru_msgsnd    0
> ru_msgrcv    0
> ru_nsignals  0
> ru_nvcsw     1382558
> ru_nivcsw    441419
> cpu          2171.596
> mem          73.691
> io           272.204
> iow          0.000
> maxvmem      223.262M
> arid         undefined
> 
> 
> I'm surprise that the wallclock time (1761) is less than the cpu time (2171) , in a single core jobs. In this case, i've got an eficiency more than 120 % !!!
> 
> I'm sure i'm wrong, but can't know where..
> 
> Thanks.
> 
> -- 
> -- Jérôme
> Si le problème a une solution, il ne sert à rien de s'inquiéter.
> Mais s'il n'en a pas, alors s'inquiéter ne change rien.
>    (Proverbe tibétain)
> _______________________________________________
> users mailing list
> users at gridengine.org
> https://gridengine.org/mailman/listinfo/users




More information about the users mailing list