End To End Web Service Recommendation Methods Oriented On Using Enviroment

As the number of available Web Service (WS) providers increases rapidly, redundancy has become prevalent with many service providers offering same or similar services. Quality of Service (QoS) is one of the most important factors when choosing a WS provided by a specific WS provider. Currently performance measurements are mostly gathered from the server-side. However, the importance of client-side performance measurement cannot be neglected, as the measurement represents the effective performance experienced by the client. As a result, WS recommendation only based on server-side performance measurements can not assure the effectiveness of the method. How to select the best WS based on different application environment of end users is a hot research topic.To solve this problem, the main issue of this thesis is:How to select the best suited service and recommend it to users of different application environment with WS functional requirements satisfied. The thesis takes full account of the different using environment factors, developes an integrated framework of WS monitoring, analysing, evaluating and recommending. In this thesis the recommended methods, which are based on end-to-end monitoring of WS and performance experienced by WS clients, can overcome the drawbacks of the unilateral trust of service providers (UDDI) or other third-party WS monitoring methods. This thesis focuses on performance which is one of the most important QoS attributes, and establishes QoS indicator system from client point of view since the QoS indicator system from client point of view are more in line with the actual needs.Based on the research of performance experienced by WS clients, the thesis focuses on the using enviroment oriented end to end web service recommendation methods, which contain three main parts:end-to-end monitoring, client grouping and service recommending. By analysis on client grouping, we present a clustering algorithm and an inference algorithm. Finally, a case study is presented to express the validity of these methods.