Presenter: John T. Murphy

Title: Agent-Based Modeling on Parallel and High-Performance Systems: Theory and Practice

Abstract: As agent-based models become larger in scale they can easily exceed the capabilities of desktop computers. Larger systems, including the high-performance systems that are now the largest computers in the world, are generally parallel environments. But running an agent-based model in a parallel environment requires that the model be divisible into semi-independent parts; the pattern of dependencies among the elements of the model can be more or less compatible with this requirement, and some agent-based models effectively cannot be parallelized. This talk will begin with a theoretical discussion of the challenges and limitations of putting ABMs in a parallel environment. The remainder of the workshop will be a hands-on, nuts-and-bolts tutorial on Repast HPC, a software package being developed at Argonne National Laboratory that extends the paradigm of the Repast Simphony toolkit to a high-performance environment and provides structures and tools that automatically deal with many of challenges of parallelizing ABMs.

Bookings