Finite-difference time-domain (FDTD) analysis using distributed computing

This paper describes an implementation of the finite-difference time-domain (FDTD) calculations using PVM [Parallel Virtual Machine] 3.2 and a cluster of workstations for a test problem involving a three-dimensional rectangular cavity. It is found that the speedup factors achieved in FDTD calculations can approach the maximum values for large problem sizes if the computation-to-communication ratios are maintained at values significantly greater than unity. It is also found that in order to achieve linear speedups, the problem sizes should be increased with the number of processors while adequate computation-to-communication ratios are maintained in the individual processors. The results demonstrate the potential of parallel distributed computing for FDTD calculations in electromagnetics.<<ETX>>