Supercomputing Applications of the Other Kind: Real-time Parallel Discrete Event Simulations of Large-scale, Smart Infrastructures

Abstract

Ultra-scale supercomputing hardware is a reality, reaching peta-scale recently and now moving to exa-scale. A rich class of applications has however remained largely untapped to reap supercomputing benefits, namely, parallel discrete event simulations (PDES), partly due to technical challenges, and partly awaiting compelling applications. The recent emergence of new visions of a ‘smarter’ and more agile societal operation has clearly opened new large-scale applications that are directly formulated as grand-scale PDES scenarios executed in an on-line, real-time fashion. Our presentation will focus on highlighting the potential for grand-scale real-time solutions, illustrated using our preliminary efforts in four example application areas: (a) state- or regional-scale vehicular mobility modeling, (b) country- or world-scale epidemic modeling, (c) cyber infrastructure and security operations at the scale of multiple autonomous systems, and, (d) country- or world-scale social behavioral modeling. We believe the context is ripe to envision and formulate similar grand-scale solutions in many additional application areas. The technical vision to progress to such an ambitious goal, however, is highly challenging. We outline some of the salient technical issues, challenges, and potential solution directions in meeting the scale and speed demanded by this ‘other kind’ of supercomputing applications.