A simple dynamic load-balancing scheme for parallel molecular dynamics simulation on distributed memory machines

Abstract

We propose a simple and efficient load-balancing scheme for parallel molecular dynamics simulation on distributed memory machines. It decomposes spatial domain of particles into disjoint parts, each of which corresponds with a processor and dynamically changes its shape to keep about the same number of particles throughout the simulation. In contrast to other similar schemes, ours requires no long-distance inter-processor communications but only those among adjacent processors (thus little communication overheads), whereas it still guarantees fast reduction of load-imbalance among the processors. It owes these advantages mainly to the following features: (1) The sufficiently correct global load information is effectively obtained with step-wise propagation of appropriate information via nearest neighbor communication. (2) In addition to the global load-balancing, another load-balancing procedure is also invoked on each processor without global load information in order to suppress rapid increase or decrease of loads. Thus, informations from remote processors can provide reliable values even after a certain period of delay. Further, we discuss how to select loads to migrate among processors so that spatial locality of the processors may be preserved. Through preliminary evaluation on an uniprocessor workstation, we have shown the scheme has strong potential for large-scale parallel molecular dynamics simulation on distributed memory machines or workstation clusters.