Hi,

Am 04.11.2011 um 10:10 schrieb Johan Finstadsveen:

> Thanks for a quick reply.
> 
> You have summarized correctly. So you would recommend a setup where the 
> gpu-frontend-node would manage all machines, cpu and gpu? That is good to 
> know, but requires some rethinking of our current setup.

I can second this.


> The desire from the users is to be able to send different workloads to 
> different queues,

SGE way of scheduling is not to send jobs to different queues by the user. The 
main idea is to request resources, and SGE will select an appropriate 
queue/exechost for your job which will fulfill the resource request. You could 
set up two complexes of type BOOL like "gpu", attach it to certain 
queues/exechosts and request `qsub -l gpu job.sh` and it will be scheduled only 
to machines where this can be satisfied.


> depending on the type of task, ie some tasks are more ideal for cpu, others 
> gpu. Additionally they wish to utilize the cpus on the gpu-nodes to maximize 
> the overall utilization. I am not sure whether this is possible, or if it is 
> an overall different debate. 
> 
> The desire from me as a sysadmin is to have the all-function used as a method 
> of removing machines temporarily to perform tests or upgrades without users 
> adding more jobs. 

For this it's possible to disable the queue instance residing on a particular 
exechost to drain it and then do some maintenance. Or have in this case two 
queues on the machines with different ACLs (i.e. one only for admin staff).

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

Reply via email to