Simplifying parallel and distributed simulation with the DUP system

This paper presents how the DUP System, a straightforward POSIX-compatible framework that enables programming-language-agnostic parallel and distributed stream processing, can be used to facilitate parallel and distributed simulations. Specifically, we describe two ways of using DUP to utilize available resources for efficient simulation: (1) a straightforward technique for parallelizing multiple runs of an existing simulation program with minimal changes, and (2) FiDES, a Discrete-Event Simulation (DES) framework built atop DUP that provides a simple, yet powerful, means of implementing a parallel and/or distributed DES. We then describe a toolset for profiling, debugging and visualization that aids the development of DUP simulations. To support these claims, we present various performance benchmarks that collectively demonstrate how DUP and FiDES can make high-performance simulation accessible to everyone.