Automated Load Balancing of a Missile Defense Simulation Using Domain Knowledge

Many simulation models require an unsatisfactory amount of time when run on a single processor. Parallelizing these simulations provides the requisite computational resources to achieve the desired running time, but also introduces communications costs that can outweigh the gains of parallelization. Load-balancing algorithms attempt to find the optimal tradeoff between placing simulation objects together to reduce communication costs and distributing them to gain parallelism. Static load-balancing algorithms assign simulation objects to processors before the Automated Load Balancing of a Missile Defense Simulation Using Domain Knowledge