Research Of Consumer Driven Web Services Dynamic Composition Method

Web Service Composition is a hotspot in SOA (Service Oriented Architecture) area . However, SOA organizes services in a"provider- centric"manner, which brings a critical problem: More services come into being, more complex and difficult the services discovery and subscription become. A"consumer driven"approach, based on user ralationship reasoning and QoS negotiation, is proposed to addressing this problem.First of all, the paper introduced a conception,"User Club", into SOA, in order to extend SOA and attach importance to user ralationship, utilizing the idea of model reuse in personalized field."Service Pool"was born in the"User Club"by clustering topics of consumer requirements. On the basis of extended SOA model with user ralationship ontology, the paper proposed a supporting framework for Consumer Driven Web Services Composition.Secondly, the"User Club"set up an index by resusing optimal service or composition service.Then, the thesis advanced a method for services aggregation, according to users ralationship ontology, which was split into three stages: first step, calculating semantic distance between topics of consumer requirements, second step, reasoning users ralationship with semantic, finally,"User Club"found candidate services list.Thirdly, WS-Policy language is adhibited to the Web service description by way of extending it in existing UDDI, and a QoS negotiation approach is proposed to form a spectrum of QoS so that consumer's quality requirements can be satisfied by automated QoS negotiation, based on WS-Policy ontology, a matching resoning is executed between the consumers'requirement assertions and the services'capability assertions.In the end, a services composition engine is implemented to information query on the basis of the supporting framework for Consumer Driven Web Services Composition. The engine will act as a service agent so that consumers only discover and subscribe the service pool instead of a large number of candidate. Experimental results show that this method has a good performance in term of search space, response time and precision than the previous when the repository has a large number of consumers and services.