Comparison of two methods of parallelizing GEANT4 on beowulf computer cluster

Numerical techniques for problem solving have become an integral part of theoretical science and mathematical modeling. However, the rate of improvement of computer processor speeds have begun to plateau. Parallel and distributed computing have become the standard method to perform computationally expensive calculations. The beowulf cluster, a group of computers connected in a local network, is a simple means of creating a powerful machine for parallel computing. We describe the setup of GEANT4 simulation software on a beowulf cluster using the ROCKS cluster operating system and test two methods of distributing work throughout the cluster: a “frontend-compute node” method using TOP-C and a “static scheduling” method using MPI.