Research On Mobility-aware Service Composition With QoS Correlations And Temporal Constraints

The development of Web service technology makes it possible to compose the existing services to fulfill more complex business requirements. It’s essential that both functional requirements and QoS should be satisfied in the composition process. This process is called QoS-aware service composition. From mobile perspective, the mobile devices can not only provide but also consume services.From the perspective of the mobile device as a service consumer, automatic ser-vice composition is significantly important in the case that a user uses the mobile device to compose services when there is not a predefined process template. Most of exist-ing automatic service composition approaches assume that services are independent of each other regarding their QoS values. However, in real setting, QoS correlations between services do exist and affect the QoS of the composite service. In this case, traditional approaches may fail to get the optimal solution.From the perspective of the mobile device as a service provider, a growing body of software applications are provided on mobile devices in the form of mobile services, which can be used for service selection. In mobile environment, different users may have different preferences on the QoS attributes. Once the preferences change, we should reselect proper service from the whole candidates and this is often inefficient, thus computing composite service skyline (CSKY) is significantly important. Mean-while, the availabilities of mobile services are associated with temporal constraints. In this case, traditional approaches may generate infeasible CSKY which violates tempo-ral constraints.These issues bring new challenges to the service composition. In this paper, we investigate these issues and propose solutions respectively. Specifically,1. For the problem of automatic service composition with QoS correlations, we propose an approach based on graph search. At first, we introduce a preprocess-ing algorithm to address the QoS correlations and an optimization technique to reduce the size of graph. Then, we present a graph search algorithm to traverse the graph in order to find the optimal composite service.2. For the problem of computing the CSKY with temporal constraints, we present a pruning-based approach. Meanwhile, a constraint optimization model is pro-posed to determine the feasibility of a composite service.3. Extensive experiments are conducted to evaluate the effectiveness and efficiency of our methods.