A Model Reduction Approach for Improving Discrete Event Simulation Performance

Discrete event simulation (DES) performance is a crucial factor when applying large scale simulation models. It limits the ability to produce high quality simulation results within given time bounds, and thus limits the ability for iterative model evaluation. In a holistic hierarchical modeling approach, models are built from basic building blocks, which define the model granularity. Hence, the resulting model granularity is not optimal with respect to the execution semantic, and thus has a limiting impact on simu-lation performance (granularity gap). The Purpose of this article is to propose an automated model reduction approach to close the granularity gap. The key idea is to merge basic model entities in order to lower model granularity and improve simulation per-formance. The article discusses requirements and limitations of model reduction in DES and presents the architecture of a re-search prototype.