Ability to introduce event or events in an evolving simulation at runtime provides a platform to study or evaluate the temporal effects of an individual event or cascading events. Such a capability can be extremely helpful in the evaluation of what-if scenarios for contingency planning and/or to determine event timings resulting in maximum gain or impact. Simulation Cloning is a technique that enables parallel execution of many logical instances of wide-ranging what-if scenario evaluating simulation instances that physically share the computation load and memory resources at runtime. Orders-of-magnitude in computation, memory and energy savings can be obtained from such cloned executions. We developed CloneX, a simulation cloning framework to which different types of simulation applications can be interfaced. With CloneX, we ascertained theoretically derived simulation cloning benefits on partial-differential equation, ordinary-differential equation and agent-based simulations. In this talk, we will introduce the simulation cloning method, discuss the CloneX framework details and touch upon the future work in this direction.
Biography: Dr. Yoginath is a research scientist in Systems and Decision Sciences group in the MIC section. He obtained his PhD in Computational Science and Engineering from Georgia Institute of Technology in 2014. His research predominantly focuses on parallel and distributed computing algorithms and methods to address the performance problems in simulations and machine learning. His research contributions encompass a range of domain science and engineering disciplines, including climate science, nuclear science, biological science, transportation networks, smart grid, and cyber physical systems.
Last Updated: March 11, 2021 - 8:22 pm