PARALLEL-PROCESSING ON HETEROGENEOUS NETWORKS FOR GIS APPLICATIONS

The use of network-based parallel computing is gaining an increasing p opularity for different reasons. Its exploitation depends on the avail ability of simple but effective methodologies to parallelize applicati ons, and the availability of portable and efficient communication libr aries to develop parallel programs. These two items are necessary to o btain performance advantages, and to ensure software portability and r eusability. In this paper we present our experience in parallelizing, in a systematic way, a class of Geographical Information Systems appli cations. We discuss the use of two well-known communication libraries (PVM and Linda). Performance results are also reported.