Load Balancing Schemes for Large Synthetic Population-Based Complex Simulators

Abstract

We present our development of load balancing algorithms to efficiently distribute and parallelize the running of large-scale complex agent-based modeling (ABM) simulators on High-Performance Computing (HPC) resources. Our algorithm is based on partitioning the co-location network that emerges from an ABM’s underlying synthetic population. Variations of this algorithm are experimentally applied to investigate algorithmic choices on two factors that affect run-time performance. We report these experiments’ results on the CityCOVID ABM, built to model the spread of COVID-19 in the Chicago metropolitan region.