The Method To Acquire Web Services QoS Based On Execution Log

Web services technology, whose application becomes more and more wide, draws more and more attention by the industry and academia, which leads to the emergence of lots of Web services with the same or similar functions. In the function-driven Web services composition, QoS which representing non-functional attributes, is of the same importance of functional attributes, because services which do not meet the requirement of QoS, just as ones which can not meet functional requirements, can not meet users'expectation. It's necessary to judge whether the service selected by the function-based requirement can meet the requirement of QoS,and providing users with the service of better QoS attributes selected from the Web services among similar functions is also very important. To carry out all of these, we have to accurately acquire the QoS attribute from the service before its execution. The results of the acquirement will be used as the judging and selecting criteria.The current Web services QoS is directly acquired through UDDI or center monitoring method. These methods have some flaws in accuracy and efficiency of the measurement. This thesis provides a Web services QoS acquiring method based on execution log. First, this method sets up an extensible QoS model based on the Web services QoS attribute set to acquire. Because the content of execution log is directly determined by the selection of the QoS attribute set and its attribute measuring method. Next, we should find out the content that should be recorded, and find a reasonable recording framework, making it works well in recording with flexibility and convenience. Then, we should acquire the QoS attribute value at the time when the service is called based on the execution data recorded in the execution log, and generate the QoS time point database. Finally, based on the time point database generated in the previous step, users could inquire the Web services historical QoS data at a certain time or during a certain period, and predict the real-time QoS data of the next time span based on the historical QoS data. The key of the acquiring method in this thesis is to obtain the QoS data of time point. It is mainly achieved by algorithm QoSEA, TSIA and LIA. They are proved to be highly efficient by analyzing the time complexity of the three algorithms. Eventually, through comparing the real QoS data with the one obtained from the log and comparing the original QoS series with the predictive data, this thesis illustrates the accuracy of the acquiring method and applicability of the predictive model.