The Selection Method Of Web Services Oriented For Adaptive Composition Services

In recent years, Web Service has been paid close attention by the industry and the academia, as a new style of distributed computing model. In practice, the single web service usually provides some single functions, which can’t meet the requirement of complex applications. Therefore, how to integrate the functions provided by various single web services to comprise a new service which could provide much more powerful function to meet the requirement of complex applications from different uses has become a hot point of web service domain. As a result of the dynamic web environment as well as the character of dynamic and autonomic of web services, the issue of the reliability of the web service has been shown out. It has become one of the criteria that assess the service selection whether the service composition has high reliability. Therefore, how to provide the services with high reliability to meet the constraint of users becomes the central issue of service selection.In the web environment, there are many web services with different QoS parameters that meet the same requirement; how to choose the services that meet the requirement of various service nodes to comprise an executable resultant service becomes a key of service composition.To address the issue mentioned above, this thesis presents a method of dynamic service selection oriented for adaptive service composition. Comparing with other service selection methods, first, the method use redundancy to confirm the reliability of service composition, which means to select many alternative services for any task in the flow rather than selecting only one service for each task; second, the method improves the mainstream QoS model, which gets the attribute information distributing various periods by analyzing the log sink of service excution rather than getting the attribute information published by the service provider. At the base of this model, the method preprocess the service sets for any task before service selection, which divides the service instances for certain task on the timeline according to their reliability. The benefit of this is that we can get the global optimum for the high reliability of service composition based on getting the local optimum for the high reliability of single service, while for the follow-up service monitoring, service substitution, dynamic adjustment and optimization stages are quite good. To address the issue of service selection with excuting time constraint, this thesis presents the service selection algorithm based on PSO reliability driver, which uses the particle swarm optimization to get the set of alternate services for each task, meeting the user constraint, with the purpose of maximum reliability of service composition. At the last, this thesis designs three experiments:effectiveness experiment, feasibility experiment, analysis of parameter adjustment experiment, which assess the web service selection method oriented for adaptive service composition presented in this thesis from different angles.