Discussion Forum

September 8, 2010 3:57pm UTC

Maximum number for groups

Hi everyone:

I have a model that I am running on a Linux 64-bit cluster, and I am trying to decrease computation time. I am new to cluster computing, so please forgive my ignorance.

I am running a parametric job through comsol batch (v.4.0a). Under the parametric settings window, there is an option to "Distribute Parameters" and limit the number of groups. This documentation notes that if the problem is too large to run on a single node, I can enable the "Maximum number of groups" so that the same parameter is solved by several nodes (I think) simultaneously.

Could someone please tell me what is meant by "Maximum number of groups"? Do I need to limit it to 1, or increase to some other value to decrease number crunching time?

And is there a better way to speed up model solver by taking advantage of multiple nodes than using this option?

Thanks for your help.

Todd

FYI - I am trying to run the model on 5 nodes with 24 processors on each, but it doesn't seem to be solving any faster that with 1 node and 24 processors.

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse