Hi,

Am 22.10.2012 um 15:38 schrieb Lukas:

> We configured on our cluster job to core binding with parallel
> environments. User can ask for multiple cores via '-pe multicore
> <number of cores>'.
> Additionally, we implemented an equal share scheduling policy [1]. The
> problem is now, that grid engines assigns each user the same number of
> running jobs, without incorporating the number of slots for
> scheduling.
> An example:
> User A, 10 jobs, single core
> User B 10 jobs, 7 cores
> Gridengine will schedule at any time the same number of jobs for each
> user,  e.g., say  4 jobs for user A and 4 jobs for user B. Meaning,
> that user B will have 7 times more cpu-cores than user A.
> Is there something (e.g. better configuration) to prevent this?

The granted slots are taken into account. I think it's a side effect of the 
reservation. You submit the parallel jobs with "-R y"?

-- Reuti


> My current configuration:
> <snip>
>> qconf -ssconf
> algorithm                         default
> schedule_interval                 00:00:10
> maxujobs                          0
> queue_sort_method                 load
> job_load_adjustments              np_load_avg=0.50
> load_adjustment_decay_time        0:7:30
> load_formula                      slots
> schedd_job_info                   true
> flush_submit_sec                  0
> flush_finish_sec                  0
> params                            none
> reprioritize_interval             00:00:00
> halftime                          72
> usage_weight_list                 cpu=1.000000,mem=0.000000,io=0.000000
> compensation_factor               5.000000
> weight_user                       0.250000
> weight_project                    0.250000
> weight_department                 0.250000
> weight_job                        0.250000
> weight_tickets_functional         100000
> weight_tickets_share              0
> share_override_tickets            TRUE
> share_functional_shares           TRUE
> max_functional_jobs_to_schedule   2000
> report_pjob_tickets               TRUE
> max_pending_tasks_per_job         300
> halflife_decay_list               none
> policy_hierarchy                  OFS
> weight_ticket                     1.000000
> weight_waiting_time               0.000000
> weight_deadline                   3600000.000000
> weight_urgency                    0.000000
> weight_priority                   0.000000
> max_reservation                   100
> default_duration                  INFINITY
> </snap>
> 
> Thank you for any hint,
> Lukas
> 
> [1] 
> http://www.gridengine.info/2006/01/17/easy-setup-of-equal-user-fairshare-policy/
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to