Towards efficient large scale epidemiological simulations in EpiGraph

We extend our MPI-based epidemic simulator with a transportation model.We simulate the spatial dynamics of the spread of influenza in Spain.We design, implement, and evaluate different performance improvement techniques.We introduce and evaluate a load-aware process-to-processor mapping algorithm. The work we present in this paper focuses on understanding the propagation of flu-like infectious outbreaks between geographically distant regions due to the movement of people outside their base location. Our approach incorporates geographic location and a transportation model into our existing region-based, closed-world EpiGraph simulator to model a more realistic movement of the virus between different geographic areas. This paper describes the MPI-based implementation of this simulator, including several optimization techniques such as a novel approach for mapping processes onto available processing elements based on the temporal distribution of process loads. We present an extensive evaluation of EpiGraph in terms of its ability to simulate large-scale scenarios, as well as from a performance perspective.