Load Balancing Strategies for Parallel Simulations on a Multiprocessor Machine†
Azzedine Boukerche, Sajal Kumar Das · 2024
Research on parallel simulation has been stimulated during the last decade by excessive computational needs of large scale applications. Such applications concern mainly the verification of the design and the performance of complex systems such as computer and communication networks, and VLSI circuits, to mention a few. Accordingly, several methods for parallel simulation have been proposed. These techniques for parallel simulation can be classified into two groups, conservative approach and optimistic approach. While conservative synchronization techniques rely on blocking to avoid violation of dependence constraints, optimistic methods rely on detecting synchronization errors at run-time and on recovery using a rollback mechanism. Despite the fact that research on parallel simulation has been ongoing for the past decade, the problem of partitioning/mapping and load balancing for parallel simulation has been the subject of limited attention.