Skip to Main Content
Scalable performance can be obtained in a network cluster of workstations by proper choice of programming paradigm and software tools such as PVM/MPI. However, the ratio of message transmission time to computation time plays a crucial role. For regular problems such as matrix computations this ratio can be easily inferred. However, for more complex problems such as evolutionary computations this ratio cannot be arrived at unless the programmer has very good problem domain knowledge.