Towards a Generalized Queuing Network Model for Self-adaptive Software Systems

A Self-adaptive Software Systems (SASSs) is composed by a managing and a managed subsystem. The former comprises system’s adaptation logic and controls the latter, which provides system’s functionalities by perceiving and affecting the environment through its sensors and actuators, respectively. Such control often conforms to a MAPE-K feedback loop, i.e. a Knowledge-based architecture model that divides the adaptation process into four activities, namely Monitor, Analyze, Plan and Execute. Performance modeling notations, analysis methods and tools, have been coupled to other kinds of techniques (e.g. control theory, machine learning) for modeling and assessing the performance of managing subsystems, possibly aimed at supporting the identification of more convenient architectural alternatives. The contribution of this paper is a generalized Queuing Network model for SASSs, where the managed subsystem is explicitly modelled, thus widening performance modeling and analysis scope to the whole system. Job classes flowing through the QN represent activities of a global feedback control loop, which is based on the system’s mode profile and implemented by class-switches operating in conformance to proper predefined class-switching and routing probabilities. Results obtained by means of a proof-of-concept addressing a realistic case study show that the generalized QN model can usefully support performance-driven architectural decision-making.