Research On Reliability Prediction Of Web Service Composition

With the progressing of IT technology, the Internet has becoming an open distributed computing infrastructure. Especially, along the emergence and popularization of Web Services, it has become the mainstream technology for service provision. In service-oriented computing, more and more corporations and organizations expect to develop complex distributed applications by the dynamic composition of Internet accessible services from independent providers. A key issue in it is the dynamic and efficient prediction of composite service's overall quality (for example, reliability) to appropriately drive the selection and assembly of services. Many research institutions home and abroad have conducted quite a few researches on reliability prediction of service composition, and have made some valuable achievements.Aiming at the deficiency of research on reliability prediction of service composition, this dissertation firstly deeply studies the architecture-based reliability prediction models and characteristic of Web Services technology, then propose an approach to analyse the predict the reliability of composite Web Service. Using the reliability prediction model, some key questions are investigated including reliability prediction of BPEL process based service composition, using service composition to implement software fault tolerance, and reliability optimization of service composition. The contributions of this dissertation are as follows:1. The characteristic of Web Services technology and the status of research on reliability prediction are analysed. Typical models and approaches of reliability prediction are evaluated.2. Aiming at the characteristic of Web Services and service composition, this dissertation proposes the composite service process net model to describe the service-usage profile of service composition based on the stochastic Petri net. The composite service process net model can be used to describe this kind of concurrent software system.3. An approach to the reliability prediction of BPEL process based service composition is proposed based on the composite service process net model. Firstly, the BPEL process is transformed to the composite service process net model, then the reliability of service composition is calculated combing with the reliabilities of component services. This dissertation also studies the sensitivity analysis approach of service composition reliability. It is used to investigate the impacts of component service's reliability and the usage profile on the reliability of composite service, and find out the key component service and scenario. 4. Software fault tolerance is an efficient way to improving the reliability of system. In service-oriented computing, software fault tolerance can easily be implemented by service composition by leveraging the redundancy intrinsic to the web service. In this dissertation, we propose an approach to implement software fault tolerance by service composition, and evaluate their reliability and performance measure.5. Using software fault tolerance in service composition will increase their cost. In this dissertation, we study the service composition's reliability optimization problem with a choice of fault tolerance strategies. A solution method based on pareto optimum solution is proposed.6. A reliability prediction tool of composite service was designed and implementioned, which can translate the BPEL process into composite service process net model, and calculate the reliability of composite service.In summary, some key questions about reliability prediction in service-oriented computing are studied in this dissertation. We believe that our contribution make a nice groundwork for the popularization of web service technology and service composition both in theory and practice.