The Research Of QoS-Aware Web Services Selection And Fault-Tolerance Of Runtime Service Composition

As the newest application model for decentralized computing,Web service is an effective mechanism for the data and service integration on the Internet. Services composition is a process that assembles multiple component services according to their functionality, semantica and logic relationships to a composite service for providing new functionality.As Web services increase explosively,the number of services with similar functionality or implementing the same function is larger and larger,so selecting the component service considering non-functionality (i.e.QoS) is a critical technology of Web service composition.In addition,the component services of composition service may fail for various factors when running in the dynamic environment.For the sake of making users aware of the failure,it is important to re-select available Web composition plans to complete the business process required,and at the same time to satisfy the users' preference and constraints set by the user or by the structure of the composite services. This is fault-tolerance of composition service at run time.Our work mainly focuses on providing business with certain quality effectively in the complex environment of Internet,according to users' requirements.First of all,a Web service architecture (Web Jet) is proposed to support QoS-aware Web service selection and failure-tolerant processing of composition services at run time. QoS model Web Jet adopted is introduced,so are the QoS computing methods of single service and composite service.Then we contrive a QoS-driven global optimization algorithms named TGA for services selection based on Genetic Algorithms (GAs).We designed a novel data model- Process Tree(PTree) model to describe combination relationships among the tasks in composition services.PTree model provides good data structure for QoS computation of specific Composition Plan,and it is also the critical factor for replanning Web Services Composition at run time.Based on PTree which can express abstract process of composite service,we design Tree-coding scheme as the chromosome code of TGA. The experiment results show that Tree-coding is effective in global optimization of services composition.We carried on a series of experiments comparing our TGA with the exact solutions for services composition(0-1 heuristics for QoS-aware Web Service Composition).The experiment results confirm that our Tree-coding Genetic Algorithms is effective in QoS-aware Web Service Composition.Due to the dynamic attributes of services composition execution environment,the component services may become unavailable or the QoS of the component service may change significantly.We researched on the run-time fault-tolerance mechanism of services composition.Three policies are proposed to adapt to dynamic environment.The policies are retrying to invoke the failure componet service, replacing the failure service with a new component service and replanning a local-process required to execute afterward. The experiments confirm the validity of replanning.For retrying and replanning mechanism,experiments are carried on to check their execution effects.