Research Of Two-stage Service Composition Algorithm Based Fuzzy Petri Net

With the development of the internet and Web Services technologies, more and more Web services with the same function appear on the internet, but these services are different in Quality of Service(such as execution time, cost). How to choose a web service to compose so that the composite service is not only satisfy the use's function demand but also excellent in Quality of Service, is a core issue in the domain of service composition.As the complement, semantic technology becomes propulsion of service composition. Semantic Web technology tries to describe web service by machine-understandable language to support automation of discovery, selection, composite, invoke and monitor of web service, It's relevant to take semantic information into account when compose web services.To solve the problem above, a formal model for Web Service composition based on Fuzzy Petri Net is given, choose composition service path by fuzzy reasoning considering service's semantic and QoS. The Work carried out in the paper is as following:Firstly, a parse method to partition a BPEL process into several sub processes based on activities associated with services is presented, a map from these sub process to Petri Net is created. These sub processes can be map into Petri Net easily, so we can model the BPEL process by Petri Net.Secondly, The semantic match model and the QoS match model of web service are created. Then Fuzzy Petri Net model for composite services can be established. Using fuzzy reasoning rule and fuzzy reasoning algorithm, the composite services'match is obtained by single service's match. Then a two-stage service compose algorithm is advanced to choose a composition service path which is excellent in QoS and Semantic match.Finally, a simulation system is built based PIPE(a simulation tool for Petri Net),. The experimental result show that Fuzzy Petri Net's fuzzy reasoning rule and the two-stage compose algorithm can solve the composition service routing problem when taking semantic and QoS into account.