In an ideal system, with a parallelizable load, the message throughput is simply the number of processors divided by the path length.