Research On Intelligent Technolgy For Reliable Web Serivces

Service-oriented computing is an advanced computing paradigm which isadopted in the distributed application systems of Internet enterprises. Its highreusability manifests that single independent services can be combined to providevalue-added services to meet the more complex functional requirements. Becauseservices are increasingly dependent meanwhile the network environment is complex,the service/service composition reliability has been studied by researchers andindustrial practitioners. To improve the reliability of the composite services as thegoal, three stages of the service composition including services selection, servicecomposition design and service composition monitoring are deeply analyzed andresearched to ensure the reliability of each stage. The research contents are as follows:First, in a services selection phase, an improved trust-aware servicesrecommendation method is proposed combining the recommendation technology andthe trust relationships whose combination methods are given. The influence onrecommendation capabilities by distrust information is taken into account. In order tobetter find similar set of users, the concept “the trusted degree” is proposed and thetrust network reconstruction method is devised to improve the recommended capacity.Second, in a services selection phase, in order to better distinguish similarservices, a novel services reputation management method is given considering notonly the users’s different preferences on the service quality attributes, but also thestatus of the service during the evaluation. In order to improve the discrimination ofservices, a user’s sensitivity is introduced. The experiment results show that theproposed comprehensive reputation model is effective.Third, in a services composition design phase, for the verification offault-handling and compensation mechanisms of the business process, we first give anextended BPEL semantic model, and then propose to use planning techniques toverify its feasibility.Fourth, in a running phase of services composition, in order to enhance itsfault-handling capability, a case-based reasoning fault-tolerant framework is proposed.Basing on summarizing various errors information, the fault-handling examples areadvised to be stroed in a database as historical cases in response to future errors. Thissection elaborates the cases’ representation, cases’ search and the reuse strategy.Meanwhile, an adaptive reconfiguration algorithm is proposed to reduce the fault-handling costs according to the domain of influence. Experiments on real dataprove the feasibility of this framework.Fifth, in order to enhance the adaptive capacity of a service composition,adaptive methods are proposed which are based on probabilistic context-freegrammars (PCFG) and learning automata respectively. They model the servicescomposition from different angles. The former concerns the interaction informationbetween services, while the latter focuses on the overall modeling including therunning environment. PCFG has good scalability on modeling the adaptive strategieswhere the step-based selection method is designed making use of probabilisticinformation. An automaton constantly updates each service’s performance by learningthrough their ability. When an error occurs, it can automatically generate the optimal(suboptimal) services composition instance. Finally, these two methods can keepbusiness processes proceeding rightly.