Regional weather prediction on small network of workstations

In this paper we present a methodology followed to parallelize a weather forecasting code running operationally on a small Network of Workstations with real data sets. The 1D Eta/NMC model code was originally written in Fortran 77 and optimized for vector supercomputers. We describe the steps followed to parallelize the Eta/NMC-dust code. The examination of program workload gave important insights into bottlenecks of the Eta/NMC-dust code. The implementation of the model with MPI libraries on a MIMD platform, the data partitioning used and some implementation characteristics of the model are presented in the paper. We present the performance of the parallel implementation of the model in terms of the speedup gained on a number of processors