A distributed implementation of the finite difference time‐domain (FDTD) method

Abstract

SUMMARYThe finite difference time-domain (FDTD) method is a well known numerical technique that has been used to solve electromagnetic boundary value problems. However, the method requires large computational resources to solve a problem, restricting its use on sequential computers to small problems. This has led to the use of expensive supercomputers in the past to solve larger problems.In this paper we first discuss existing parallel implementations of the FDTD method. We then describe an approach to compute the FDTD equations based on distributed-memory computer architectures. Finally we evaluate a low-cost parallel implementation of the FDTD method that runs on a network of workstations, utilizing spare network capacity.