Am 04.04.2013 um 11:14 schrieb Arnau Bria:

> On Tue, 26 Mar 2013 17:10:55 +0100
> Reuti Reuti wrote:
> 
>> Hi,
> Hi Reuti,
> 
> [...]
> 
>>> 2) core binding. I have it configured for serial and smp jobs, but
>>> which is the correct strategy and configuration for mpi jobs?
>>> Is linear going to span jobs acros different host sockets?
>> 
>> AFAICS the request is applied on all machines which you get granted
>> for the job. I.e. applied per `qrsh -inherit ...` besides setting it
>> for the jobscript already. This is hard to handle in case of a round
>> robin allocation, as you don't know in advance whether you get just
>> one slot per machine or more. Maybe the best would be to use it with
>> a fixed allocation rule only.
> Seems the best idea, but it has a big con: 
> 
> if I define an allocation of, let's say 2 slots per node, and there's a

Or get always full nodes, and let Open MPI do the binding.

-- Reuti


> job asking for 40 slots, I need 20 nodes to free 2 slots each, if one
> node has 12 slots free, only 2 will be used, will really affect
> adversely mpi queue jobs cause maybe some of those 20 nodes are running
> really long jobs...
> 
> I'm thinking on leaving mpi jobs out of any core binding strategy. 
> 
> 
>> -- Reuti
> Cheers,
> Arnau
> _______________________________________________
> 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