Model-driven performance prediction of HLA-based distributed simulation systems

Performance models offer a convenient tool to assess design alternatives and predict the execution time of distributed simulation (DS) systems at design time, before system implementation. Currently, performance models are to be manually developed and the related extra effort often becomes the limiting factor for their cost- and time-effective use. In this paper, we aim to reduce this extra effort with the introduction of a model-driven method for the automated building of performance models whose evaluation provides a prediction about of the execution time of a distributed simulation system. As such, the method contributes to bring software performance engineering techniques into the distributed simulation system lifecycle. In particular, we show how the SysML-based specification of the system to be simulated and the design documents of the DS system can be used to derive the topology and the parameters of a performance model specified according to the Extended Queueing Network formalism.