A Model-Driven Approach For Dynamic Web Service Composition

With the development of Web services, Service Oriented Architecture (SOA) has been widely adopted by more and more enterprises. The key issue in the success of SOA is the ability to increase business agility by dynamically composing Web services and adapting business processes to fast business changes. Web Services Business Process Execution Language (WS-BPEL), which is recognized as today's standard business process description language, provides the capability of defining executable business processes. However, current business process development is lack of dynamicity, since it only focuses on the low-level reuse of Web services. It is desirable that the business process development procedure supports the high-level reuse of business processes and provides dynamic service selection and binding capabilities.This dissertation delivers a detailed introduction of our model-driven approach for dynamic Web service composition. First, we define a platform-independent service model and a language-independent business process model based on Unified Modeling Language (UML). They are used for modeling Web services and business processes, as well as describing the semantics of Web service functionalities and business process requirements. Next, we generate Web service implementations and language-specific business processes from the two models automatically, by following the predefined mapping patterns and transformation rules. The model transformation techniques enable the service and business process models to be reused cross-platform. Finally, the Web service matchmaking and binding operations are performed based on the semantics of business process requirements and Web service functionalities. The matched Web services are dynamically selected and bound into the business processes. In this way, the Web services are dynamically composed and the business processes are better reused. The feasibility of our approach is effectively demonstrated by the prototype system presented in this dissertation.