Timed FSM strategy for optimizing web service compositions w.r.t. the quality and safety issues. (Stratégie basée sur les machines à états finis temporisées pour optimiser la composition de services web à l'égard de la qualité et de la sécurité)

Service-oriented architecture (SOA) together with a family of Everything-as-a-Service (XaaS) concepts nowadays are used almost everywhere, and the proper organization of collaborative activities becomes an important challenge. With the goal of bringing to the end-user safe and reliable service with the guaranteed level of quality, issues of service compositions verification and validation become of high practical and theoretical interest. In the related works, numerous models and techniques are proposed, but mostly focused on functional and non-functional issues in isolation, while integration of these parameters within unified formal framework still remains a problem to be solved – and therefore became one of the core objectives of this thesis. In our work, we address the problems of web service composition verification and optimization with respect to functional, quality and safety properties of the composition. Finite state models are proven to be useful for testing and verification purposes as well as for service quality evaluation at each step of service development. Therefore, we propose to use the model of Finite State Machine with Timeouts (TFSM) for integrating functional service description with time-related quality and safety parameters, and derive the extension of the model in order to adequately inherit significant nondeterminism due to the lack of observability and control over third-party component services. For the purpose of component optimization in the composition, we propose a method for deriving the largest solution containing all allowed component service implementations, based on solving TFSM parallel equation. Further, techniques for extracting restricted solutions with required properties (minimized/maximized time parameters, deadlock- and livelock-safety, similarity to the initially given component, etc.) have been proposed. In cases when the specification of a composite service is provided as a set of functional requirements, possibly, augmented with quality requirements, we propose a technique to minimize this set with respect to the component under optimization. Application of the obtained results for more efficient candidate component services discovery and binding, alongside with extending the framework for more complex distributed modes of communications, are among the topics for the future work.