[GE users] Slots limitation ....

TRAN Chanh chanh.tran at dassault-aviation.fr
Mon Apr 4 09:14:56 BST 2005


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi Reuti,

Based on your tips, I finally figured out the formula which establishes 
the limitation on slots :

nb_slots = ceil(2 * nb_proc *  np_load_average)
- np_load_average = 1.75 & nb_proc = 1 => nb_slots is limited to 4
- np_load_average = 1.75 & nb_proc = 2 => nb_slots is limited to 7
and so forth ....

Actually, I set, for this specific queue, np_load_average = INFINITY & 
it works :-) ...

Thanks again,
Chanh

Reuti wrote:

>Hi Chanh,
>
>I asked about the evaluation installation as I see no speed benefit running a 
>parallel job on a single CPU machine. Having 8 slots for serial jobs to allow 
>some kind of time sharing between users is okay, although each job will run 
>longer of course.
>
>Back to your problem: 7 is just the "load_thresholds      np_load_avg=1.75" 
>times 4 it seems.
>
>Can you try setting it to NONE? As you don't honor the np_load_avg now, you can 
>also optimize the scheduler by setting with "qconf -msconf":
>
>job_load_adjustments       NONE
>load_adjustment_decay_time 0:0:00
>
>Cheers - Reuti
>
>
>Quoting TRAN Chanh <chanh.tran at dassault-aviation.fr>:
>
>  
>
>>Hi Reuti,
>>
>>Here 're the infos :
>>
>>Case w/ 2_proc machine 'nvb0' :
>>
>>#####################################################
>>
>>processor  : 0
>>vendor     : GenuineIntel
>>arch       : IA-64
>>family     : Itanium 2
>>model      : 1
>>revision   : 5
>>archrev    : 0
>>features   : branchlong
>>cpu number : 0
>>cpu regs   : 4
>>cpu MHz    : 1495.930995
>>itc MHz    : 1495.930995
>>BogoMIPS   : 2239.75
>>
>>processor  : 1
>>vendor     : GenuineIntel
>>arch       : IA-64
>>family     : Itanium 2
>>model      : 1
>>revision   : 5
>>archrev    : 0
>>features   : branchlong
>>cpu number : 0
>>cpu regs   : 4
>>cpu MHz    : 1495.930995
>>itc MHz    : 1495.930995
>>BogoMIPS   : 2239.75
>>
>>
>>####################################################
>>
>>qconf -sq test.q
>> ==> SGE version 5.3p6
>>qname                test.q
>>hostname             nvb0.dassault-avion.fr
>>seq_no               0
>>load_thresholds      np_load_avg=1.75
>>suspend_thresholds   NONE
>>nsuspend             1
>>suspend_interval     00:05:00
>>priority             0
>>min_cpu_interval     00:05:00
>>processors           UNDEFINED
>>qtype                BATCH INTERACTIVE PARALLEL
>>rerun                FALSE
>>slots                10
>>tmpdir               /tmp
>>shell                /bin/csh
>>shell_start_mode     NONE
>>prolog               NONE
>>epilog               NONE
>>starter_method       NONE
>>suspend_method       NONE
>>resume_method        NONE
>>terminate_method     NONE
>>notify               00:00:60
>>owner_list           NONE
>>user_lists           NONE
>>xuser_lists          NONE
>>subordinate_list     NONE
>>complex_list         NONE
>>complex_values       NONE
>>calendar             NONE
>>initial_state        default
>>s_rt                 INFINITY
>>h_rt                 INFINITY
>>s_cpu                INFINITY
>>h_cpu                INFINITY
>>s_fsize              INFINITY
>>h_fsize              INFINITY
>>s_data               INFINITY
>>h_data               INFINITY
>>s_stack              INFINITY
>>h_stack              INFINITY
>>s_core               INFINITY
>>h_core               INFINITY
>>s_rss                INFINITY
>>h_rss                INFINITY
>>s_vmem               INFINITY
>>h_vmem               INFINITY
>>
>>#########################################################
>>
>>qconf -sp test
>> ==> SGE version 5.3p6
>>pe_name           test
>>queue_list        test.q
>>slots             10
>>user_lists        NONE
>>xuser_lists       NONE
>>start_proc_args   /bin/true
>>stop_proc_args    /bin/true
>>allocation_rule   $pe_slots
>>control_slaves    FALSE
>>job_is_first_task FALSE
>>
>>#########################################################
>>
>>#!/bin/ksh
>>
>>
>>#$ -S /bin/ksh
>>#$ -o /tmp/test.out
>>#$ -j y
>>
>>#$ -pe test 8
>>
>>sleep 60
>>
>>#########################################
>>
>>works only w/ nb_slots < 8 !!!
>>
>>#######################################
>>
>>I don't know what you mean by "evaluation installation for parallel 
>>programs",
>>matter-of-factly my SGE has been set up for quite a while already & has 
>>got no pb running // jobs.
>>The new thing is up to now I always have 'nb_slots = nb_procs' .....
>>
>>CU,
>>Chanh
>>
>>Reuti wrote:
>>
>>    
>>
>>>Hi,
>>>
>>>do you request any resources by default or explicitly for your jobs (memory,
>>>      
>>>
>>>licenses,...)? Please post your queue/PE definition and how you submit (is
>>>      
>>>
>>it 
>>    
>>
>>>an evaluation installation for parallel programs?).
>>>
>>>CU - Reuti
>>>
>>>Quoting TRAN Chanh <chanh.tran at dassault-aviation.fr>:
>>>
>>> 
>>>
>>>      
>>>
>>>>Hi all,
>>>>
>>>>I'm using 5.3p6 & my pb is :
>>>>
>>>>- I 've  :
>>>>   a machine w/ one single processor,
>>>>   a queue named 'single.q' w/ 8 slots,
>>>>   a PE named 'pe_8'
>>>>
>>>>With all the 8 slots available, I can submit jobs requesting up to 4 
>>>>slots but not > 4
>>>>w/ msg "cannot run because resources requested are not available for 
>>>>parallel job"
>>>>
>>>>The same thing w/ a 2_processor machine & the limit this time is 7 
>>>>instead of 4.
>>>>
>>>>Is there a limitation on the number of slots  in regard to the  number 
>>>>of  processors ???
>>>>
>>>>Thanks in advance for any explanations,
>>>>Chanh
>>>>
>>>>
>>>>---------------------------------------------------------------------
>>>>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>>>For additional commands, e-mail: users-help at gridengine.sunsource.net
>>>>
>>>>   
>>>>
>>>>        
>>>>
>>>
>>>---------------------------------------------------------------------
>>>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>>For additional commands, e-mail: users-help at gridengine.sunsource.net
>>>
>>>
>>> 
>>>
>>>      
>>>
>>---------------------------------------------------------------------
>>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>For additional commands, e-mail: users-help at gridengine.sunsource.net
>>
>>    
>>
>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>
>  
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list