Hi, Am 19.04.2013 um 11:43 schrieb Sangamesh Banappa:
> So it is 10 * 16 = 160 hours of WALLTIME. > > I still have doubt on this. Because: > > The cluster is configured with two queues: > > (1) serial.q with 20 cores (2) parallel.q with 224 cores. > > But according to the GE accounting data, usage time of both queues is > almost same. > > total_usage = 179283327 > serial_usage = 89922184 > parallel_usage = 89361143 For parallel jobs the wallclock time still stays the same, being it one or whatever amount of corses used. Initially I referred to serial jobs only in a fully loaded cluster and misread your last post. Sorry for not being precise. -- Reuti > Any hints for this? > > Thanks, > Sangamesh > > > ----- Original Message ----- >> Am 19.04.2013 um 06:12 schrieb Sangamesh Banappa: >> >>> Hi Reuti, >>> >>> Thanks for the quick reply. >>> >>> So the WALLCLOCK is the accumulated wallclock time of the jobs >>> which ran. >>> >>> To be more clear, if a job has run on 16 cores for about 10 hours, >>> then WALLTIME is just 10 hours or 10 * 16 = 160 hours? >> >> Exactly. - Reuti >> >> >>> ----- Original Message ----- >>>> Am 18.04.2013 um 14:30 schrieb Sangamesh Banappa: >>>> >>>>> Hi, >>>>> >>>>> I need some details on the accounting data captured by >>>>> GridEngine. The qacct output for the last 90 days is: >>>>> >>>>> # qacct -d 90 >>>>> Total System Usage >>>>> WALLCLOCK UTIME STIME >>>>> CPU MEMORY >>>>> IO IOW >>>>> ================================================================================================================ >>>>> 179283327 1188903707.397 11455939.791 >>>>> 1624833514.575 760640736.875 16275.371 >>>>> 0.000 >>>> >>>> This is just the accummulated wallclock time of the jobs which >>>> ran. >>>> It can even be higher than the number of cores multiplied by the >>>> time in case more slots than cores were defined and used. >>>> >>>> -- Reuti >>>> >>>> >>>>> This cluster has 16 Nodes (16 cores per node) and its installed >>>>> in >>>>> September 2012 with Rocks. And was down for 1 month. >>>>> >>>>> I'm not getting how these metric values reflected here. >>>>> >>>>> Hoping that these are in seconds:- >>>>> >>>>> The WALLTIME = 179283327 seconds >>>>> >>>>> >>>>> 1 hr = 3600 seconds >>>>> 179283327 / 3600 = 49800.924166667 hours >>>>> >>>>> 1 day = 24 hours >>>>> 49800.924166667 / 24 = 2075.038506944 days >>>>> >>>>> Number of Nodes = 16 >>>>> >>>>> 2075.038506944 / 16 = 129.689906684 days of usage per Node >>>>> >>>>> Number of cores in each Node = 16 >>>>> >>>>> 129.689906684 / 16 = 8.105619168 days of usage per core? >>>>> >>>>> Is this calculation correct? >>>>> >>>>> Can someone give more details about these metrics? >>>>> >>>>> Thanks in advnace >>>>> _______________________________________________ >>>>> users mailing list >>>>> [email protected] >>>>> https://gridengine.org/mailman/listinfo/users >>>> >>>> >> >> _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
