Quality Failure Prediction for the Self-Healing of Service-Oriented System of Systems

With software systems increasingly being employed in more complex and critical contexts, service-oriented system of systems has been paid more and more attention as a novel software system structure, which considers System as a Service. Under the loosely coupled SoS's dynamic and uncertain running environment, self-healing process, as the important safeguard mechanism of system running, pose a great threat to system quality analysis. Particularly, as the first step of self-healing, the research of quality failure prediction faces not only continual and immediate disturbance on service quality, but also complex users' preference on quality. In this paper, we propose a model based on Stochastic Differential Equations to analyze the disturbance more precisely and dynamically. And we adopt weighted conditional preference to consider different users' requirements. This model is testified in an empirical case study, in which the real data set is collected in real-time from the system platform of a Telecom in China. The experiments verify model's prediction abilities and evaluate the impact of the parameters on the prediction accuracy.