اگه انگلیسیتون خوبه احتمالا این کمکتون کنه:

If you are considering only a single node with multiple processor then multi-threading is the best way, also the performance is better. And for clusters i.e running a program to span across multiple nodes then MPI is the probably the best way. Also you can embed multi-threading within each MPI processes on these nodes which are running given MPI program.