Towards in situ visualization of extreme-scale, agent-based, worldwide disease-spreading simulations

Exascale computing is regarded as a major milestone for a number of application fields. However, utilizing the massively parallel power of such architectures requires a significant paradigm shift in the software structure of both simulation and visualization systems. A suboptimal workload distribution, communication bottlenecks between processes, as well as global memory accesses are often the limiting part in current approaches when it comes to the execution of such software in an exascale environment. Recent cases of malaria and Ebola in Africa have shown that there is a need for disease-spreading simulations to help healthcare workers better predict unexpected scenarios and take the necessary precautions. Reaction time plays an important role when it comes to containing a disease. Therefore, converting these kinds of demanding simulations to highly parallel computing systems as well as interactively visualizing the results is key. This paper demonstrates an approach towards a worldwide simulation of disease-spreading and an integrated---in situ---visualization. Instead of applying large-scale stochastic models, our agent-based method simulates the complete population of the world, accounting for daily tasks and routines of each subject. We focus on massive parallelization and efficient scaling to benefit from existing petascale to upcoming exascale setups. In particular, workload distribution is performed based on space partitioning and not on the population count, which is nonetheless included as an initial optimization. The resulting simulation is combined with a query-driven in situ visualization client that directly utilizes the spatial partitioning and the employed hierarchy. Our preliminary benchmarks underline the scalability of our approach in terms of rising computing power and encourage further research in the direction of worldwide, agent-based simulations and the corresponding in situ visualizations.

[1]  Shawn T. Brown,et al.  FRED (A Framework for Reconstructing Epidemic Dynamics): an open-source software system for modeling infectious diseases and control strategies using census-based populations , 2013, BMC Public Health.

[2]  James P. Ahrens,et al.  An Image-Based Approach to Extreme Scale in Situ Visualization and Analysis , 2014, SC14: International Conference for High Performance Computing, Networking, Storage and Analysis.

[3]  Patrick Weber,et al.  OpenStreetMap: User-Generated Street Maps , 2008, IEEE Pervasive Computing.

[4]  John J. Grefenstette,et al.  FRED Navigator: An interactive system for visualizing results from large-scale epidemic simulations , 2012, 2012 IEEE 8th International Conference on E-Science.

[5]  Alessandro Vespignani,et al.  The GLEaMviz computational tool, a publicly available software to explore realistic epidemic spreading scenarios at the global scale , 2011, BMC infectious diseases.

[6]  Cláudia T. Codeço,et al.  Epigrass: a tool to study disease spread in complex networks. , 2007, Source Code for Biology and Medicine.

[7]  L Ramírez-RamírezLilia,et al.  A new surveillance and spatio-temporal visualization tool SIMID , 2013 .

[8]  Todd D. Ringler,et al.  Climate modeling with spherical geodesic grids , 2002, Comput. Sci. Eng..

[9]  Dennis L. Chao,et al.  FluTE, a Publicly Available Stochastic Influenza Epidemic Simulation Model , 2010, PLoS Comput. Biol..

[10]  Madhav V. Marathe,et al.  EpiSimdemics: An efficient algorithm for simulating the spread of infectious disease over large realistic social networks , 2008, 2008 SC - International Conference for High Performance Computing, Networking, Storage and Analysis.