It is important to scale the simulation size with the number of cluster nodes being used. The optimal scaling is determined from a ratio of the time to compute and the time taken to exchange data between cluster nodes. If the simulation is distributed between too many cluster nodes, the time taken to perform the necessary data exchange will dominate the total time taken to compute each simulation iteration; consequently, no benefit will be achieved from using additional cluster nodes.
The time taken to exchange data between cluster nodes is not only determined by the size of the simulation and the number of cluster nodes used, but also by the speed at which the network hardware can exchange data.
Therefore, the optimal number of cluster nodes to use is dependent upon the hardware of the cluster and is difficult to state a priori. It is recommended that when using 100/10BaseT Ethernet cards, the minimum number of mesh cells placed on each cluster node is 100,000, whereas on higher-performance network hardware this limit can be decreased.