Massively parallel simulations of spread of infectious diseases over realistic social networks

Controlling the spread of infectious diseases in large populations is animportant societal challenge. Mathematically, the problem is best captured as acertain class of reaction-diffusion processes (referred to as contagionprocesses) over appropriate synthesized interaction networks. Agent-basedmodels have been successfully used in the recent past to study such contagionprocesses. We describe EpiSimdemics, a highly scalable, parallel code writtenin Charm++ that uses agent-based modeling to simulate disease spreads overlarge, realistic, co-evolving interaction networks. We present a new parallelimplementation of EpiSimdemics that achieves unprecedented strong and weakscaling on different architectures-Blue Waters, Cori and Mira. EpiSimdemicsachieves five times greater speedup than the second fastest parallel code inthis field. This unprecedented scaling is an important step to support the longterm vision of real-Time epidemic science. Finally, we demonstrate the capabilities of EpiSimdemics by simulating thespread of influenza over a realistic synthetic social contact network spanningthe continental United States (~280 million nodes and 5.8 billion social contacts).