Research On Dynamic Composition Of Web Services And QoS Optimization With Uncertainty

Web services composition is the task of combining a set of single functional Web services together to create a more complex and cross-organizational composite service.It is one of the most crucial technologies in Service-Oriented Computing.It is the core method to realize the reuse and value-added of services.The uncertainty of Web service is the most important inherent characteristic due to its deployment and invocation within a real and highly dynamic Internet environment.Therefore,how to solve uncertain Web services composition and composite QoS optimization problem is becoming a big research challenge.Some effects on Web services composition have been made by researchers in recent years.However,most of them focus on certain Web services composition and less consider uncertain features in dynamic environment,including non-deterministic effects in functional aspect and uncertain QoS in non-functional aspect.Therefore,for the research problems above,this thesis mainly focuses on the research of dynamic Web service composition and QoS optimization with uncertainty.The main work and contributes are as follows:(1)Analyzed the current research status of Web services composition with uncertainty.Then,proposed the formal model and description of uncertain Web service in both functional and non-functional aspects;On the basis of that,gave the problem definitions of uncertain Web service composition problem and the composite service optimization problem with uncertain QoS.Thus,they provided the basic theoretical support for the research methods and technical theory of this thesis.(2)On functional aspect,to solve the problem of non-deterministic effects in U-WSC,the thesis proposed a dynamic Web service composition approach based on non-deterministic planning conversion.Firstly,a U-WSC problem was modeled into a fully observable non-deterministic planning problem.Then,some transition rules and algorithms were proposed to translate Web service repository and service request by WSDL and WS-BPEL into non-deterministic planning domain and domain problem by PPDDL.Finally,the transformed U-WSC planning problem was solved by a highly efficient state-of-the-art uncertain planner and the experimental results proved that the approach of model and transition for U-WSC problem was feasible and effective.(3)On functional aspect,to solve the problem of some uncertain Web services with the same functionality existing in a service repository,the thesis proposed a method for U-WSC based on and-or graph planning.Firstly,the approach still converted a U-WSC problem into a non-deterministic planning problem.Then,two heuristic graph search algorithms based on LAO* and BFS algorithm,named UCLAO* and BHUC,were proposed to solve the converted uncertain planning problem.Finally,through the compare and analysis,the approach not only could find all possible uncertain composite solution paths satisfied user's request but also could effectively reduce the state search space and improve the search efficiency.(4)On non-functional aspect,to find the optimal composite service in Web service composition with uncertain QoS,the thesis proposed an optimal approach for uncertain QoS-aware Web services composition problem based on workflow model.Initially,the uncertain QoS was expressed by matrix and interval number,that realized the non-functional modeling for Web service.Then,an uncertain QoS-aware Web services composition problem was translated into an interval number-based multi-objective optimization problem with global constraints.Finally,a new improved non-deterministic multi-objective evolutionary algorithm based on decomposition was proposed and found all the optimum composite services for user's multi-dimensional preferences.