Adaptive REST Applications via Model Inference and Probabilistic Model Checking

In this paper we present a novel approach for adaptive REST Web applications that focuses on adaptation against changes in the navigational behaviour of users. The proposed solution exploits the Web server’s log file to infer a Markov model that captures the navigational behaviour of system users over time probabilistically. The model is inferred incrementally as soon as new requests are issued to the server, and is analysed periodically to verify quantitative properties by means of probabilistic model checking. The results of the run-time verification trigger ad-hoc adaptation policies, which adjust the application to the user behaviours captured by the inferred model. The paper discusses the advantages of adopting probabilistic model checking for Web applications in terms of incrementality, retroactivity and efficiency, and illustrates these characteristics as well as the applicability of the approach with a practical example.