Dynamic load balancing for large-scale multiphysics simulations

Abstract

In parallel computing load balancing is an essential component of any efficient and scalable simulation code. Static data decomposition methods have proven to work well for symmetric workloads. But, in today’s multiphysics simulations, with asymmetric workloads, this imbalance prevents good scalability on future generation of parallel architectures. We present our work on developing a general dynamic load balancing framework for multiphysics simulations on hierarchical Cartesian meshes. Using a weighted dual graph based workload estimation and constrained multilevel graph partitioning, the required runtime for industrial applications could be reduced by 40% of the runtime, running on the K computer.