Performance Study of a Minimalistic Simulator on XSEDE Massively Parallel Systems

Scalable Simulation Framework (SSF), a parallel simulation application programming interface (API) for large-scale discrete-event models, has been widely adopted in many areas. This paper presents a simplified and yet more streamlined implementation, called MiniSSF. MiniSSF maintains the core design concept of SSF, while removing some of the complex but rarely used features, for sake of efficiency. It also introduces several new features that can greatly simplify model development efforts and/or improve the simulator's performance. More specifically, an automated compiler-based source-code translation scheme has been adopted in MiniSSF to enable scalable process-oriented simulation using handcrafted threads. A hierarchical hybrid synchronization algorithm has been incorporated in the simulator to improve parallel performance. Also, a new set of platform-independent API functions have been added for developing simulation models to be executed transparently on different parallel computing platforms. In this paper, we report performance results from experiments on different XSEDE platforms to assess the performance and scalability of MiniSSF. It is shown that the simulator can achieve superior performance. The simulator can adapt its synchronization according to the model's computation and communication demands, as well as the underlying parallel platform. The results also suggest that more automatic adaptation and fine-grained performance tuning is necessary for handling more complex large-scale simulation scenarios.