[GE users] mpi & matlab

Daniel Templeton Dan.Templeton at Sun.COM
Tue May 20 19:56:12 BST 2008


    [ 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. ]

Grid Engine knows as much as you tell it.  If you configure a parallel 
environment for Grid Engine, it's able to use those nodes 
appropriately.  Depending on the nature of a MDCE job, you may either 
want to configure a parallel environment for it or just model that 
ability to run MDCE as a resource.  There are some good howto's on the 
gridengine.sunsource.net website for setting up SGE with parallel 
environments.

(From a high level, Grid Engine picks where the parallel job tasks 
should run, but it's up to the parallel environment to actually put the 
jobs there, hopefully via a special Grid Engine submit.)

Daniel

Ender Wiggins wrote:
> Hi,
>
> I'm setting up GE on a small cluster and I've got a few questions 
> about how GE works and how to plan my configuration.
>
> My nodes do MPI parallel tasks, as well as distributed task using the 
> Matlab DCE (MDCE). Each node is a SG execution host, and each node 
> runs an MDCE process as well. And I have once submit host.
>
> When a job is submitted, how does GE differentiate between what nodes 
> are MPI workers, and what nodes are capable of running MDCE. In other 
> words, how does it choose which nodes are capable and available for 
> the task? I can already run parallel task w/o using GE, but 
> integrating the two seems to be an issue that I haven't quite 
> connected yet.
>

---------------------------------------------------------------------
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