OF distributed
 

Dear all,

due to performence problems I am currently trying to use OF in distributed mode as described in the user guide. Does anybody know some numbers about how large speedup will be?

I managed to distribute data to scratch on all of the cluster nodes and run from there. Obvisously, the number of root paths in the decomposeParDict must be the same as the number of cores, even if all the nodes have the data at the same path?
I am unsure about what I need to copy back after the run? Since I copy all the processor* directories as well as system and constant (without polyMesh) to all nodes where are the results? Do I need to copy back all processor*s from all cpus? How do I know where the "correct" final solution for which processor is located in?

Thanks and regards

BastiL