Performability Modeling of Self-Adaptive Systems Based on Extension Neural Rewriting Stochastic Petri Nets

Abstract

Traditional mathematical formalisms are unable to model modern self-adaptive discrete event systems (ADES) because they cannot handle behaviors that change at run-time in response to environmental changes. This paper introduces a new extension of Reconfigurable Stoc-hastic reward Nets (RSRN), called Extension Neural Rewri-ting Petri Nets (ExNRPN), which enables the performability modeling and simulation of modern ADESs. ExNRPNs are obtained by incorporating in some special transitions of RSRNs an extension neural network (ENN) algorithm where the run-time calculation and reconfiguration is done in the local components, while the adaptation is performed for the whole system. The application of the proposed ExNRPN is illustrated by performability modeling a particular ADES.