Research On Reliability Allocation Of Web Service Composition Considering Cost Optimization

With the development of Web Service technology, the Quality of Service (QoS) has aroused wide concern in the public. QoS refers to the response time, price, availability, reliability, credibility and other non-functional attributes.Reliability and price as two important attributes of the QoS, and the reliability is an important basis of selecting Web services by service consumers, besides the price is the primary factor to consider for service provider.Some scholars has to do some research in this field, but there are still some shortcomings, such as improving the reliability only but not control the costs of Web service.This thesis studies some key problems about the reliability and cost of Web service, including quantitatived reliability, reliability constraint assigned to each component service, minimized cost of Web service composition.In this thesis, the main research is allocating the reliability constraint of the Web service composition to each of component services with minimum cost, a service composition with high reliability and low cost could be provided.This thesis proposed two methods to solve this problem, one is base on the failure rate, first of all, concluded four kinds of structure patterns of service composition based on the structured process of Web services, the Sequence, Loop, XOR, and AND patterns.And analyzed the failure modes for these patterns.Then, a method is proposed to map these failure modes to fault tree. With fault tree analysis method, we obtain the failure function of the service composition. Based on the relationship between the failure rate and the cost of the component services, we design a nonlinear programming model to allocate reliability constraint to component services reasonably.Based on reliability, this thesis proposed the other method, established the reliability allocation optimization model, using genetic algorithm to solve the model.Both methods can satisfy the reliability constraint of the service composition with minimum cost. Finally, through the contrast experiment proves the effectiveness, practicality and efficiency of the proposed method.