Dynamic partitioning of distributed virtual simulations for reducing communication load

HLA-based simulations can experience performance degradation due to communication latencies between simulation federates, which generate significant cumulative overhead. Even though the HLA standard provides mechanisms to decrease the misuse of network resources, it does not present any tool to diminish the communication latencies between interactive federates. Moreover, the interaction dependencies can be predicted before simulations are initiated, but such predictions relies on the determinism of simulations, producing erroneous balancing when simulations change their load dynamically. Thus, an hierarchical three-phase dynamic communication load balancing scheme is devised to react to run-time load changes, so the scheme performs constant, periodical monitoring of resources, re-distribution of load, and migration of federates. The balancing system reorganizes the distribution of large-scale HLA-based simulations, so the communication latencies are minimized, increasing the parallelism of the distributed simulations and leading to a performance improvement. Experiments were realized to measure the benefits of the scheme, and through comparative analyses, the balancing scheme presented considerable performance improvement to HLA-based simulations.