Performance indices to characterise concurrent applications: experimenting GSPN evaluation techniques in plant automation

This work is part of an investigation aiming at experimenting the use of Generalised Stochastic Petri Nets to model and to evaluate concurrent applications over their target parallel architectures. Former work proposed a modular methodology integrating the modelling and the performance evaluation of both sw applications and hw architectures. Its experimentation over real case studies pointed out the need to deepen some specific steps of model construction and analysis. We here collect the main considerations emerged during the experimentation activity. In particular, we discuss the experienced performance metrics and parameter assignment criteria; then we focus on the definition and use of a set of quantification indices, which have revealed to be suitable to characterise an application in terms of its performance and to support its mapping over a parallel architecture. A portion of a case study taken from ENEL R&D activity in power plant automation is used to show the usefulness, efficacy and expressiveness of the above novelties. Guided by the needs of the industrial final user to which the methodology is oriented, keywords of the whole work have been simplicity of use and flexibility, aiming at increasing practicability and (re)usability in a wide applicative area.