EpiClone: Efficient GPU-based Cloning System for Massive What-If Decision Analyses of Epidemic Models

Massive ensembles of simulations are needed to accurately capture the state space of epidemiological outbreak models. This is because the outbreak has tremendous spatial dynamism and the intervention actions have significant temporal effects. The spread of infectious diseases depends not only on the interactions among the hosts in a confined community, but also on the spatial reach of hosts among different communities. Further, intervention policies to control the epidemic may depend on various constraints such as budget, community size, movement patterns among communities. In this work, we present EpiClone, a fast and efficient spatial epidemic simulator to execute millions of different epidemic simulations at various spatial resolutions. Based on our simulation cloning framework called CloneX, EpiClone runs reaction-diffusion models of epidemiological outbreaks. EpiClone uses GPUs effectively to execute a large number of what-if decision scenarios stemming from different intervention policies. EpiClone has been benchmarked to scale up to 1024 GPUs with speedup of over two orders of magnitude compared to replicated runs.