Efficiency on clusters of parallel simulation of sine-Gordon equation over hexagonal tessellation

Experiences in the parallel execution of a numerical method to simulate the sine-Gordon equation over two different clusters are presented in this paper. The nonlinear sine-Gordon equation has been discretized in a finite differences method based on hexagonal tesselation, and then parallelized using MPI for parallel execution on three workstation clusters, a Pentium III PC based cluster using Gigabit Ethernet and Myrinet as interconnection networks, an Itanium2 based HP RX2600 cluster with Gigabit Ethernet, and a Prescott based PC cluster. Efficiency measurements over these clusters are also studied and compared.