QoS-Based Top-K Web Service Selection In Service Composition

With the development of distributed object and XML technology, Web Services comes into being. Absorbing advantage of distributed computing, Grid computing, and XML technology, Web Services highly supports interoperability, platform independent and loosely coupling, and then it turns into the main development trends of WWW Web Services provides effective means for B2B-commerce and it allows business processes as web services to be published over network by the service providers such as organizations or agencies. But the functionality of a single web service is of limitation, it is necessary to combine many single services into a composite service to make service value-added and reused. As the number of available services grows, the number of alternative services will certainly also grow, which not only increases the selection chance, but makes the selection more difficult. So in competitive environment where a large number of similar services are available, the issue of service selection becomes the focus of what we discuss and draw attention.Based on QoS, the service can be selected by comparison to users, so the QoS plays an important part in service selection. As a standard of evaluating web services, it is necessary to normalize the QoS in terms of different data types of QoS parameters. Most of the traditional selection methods are based on the QoS computation model to aggregate all the QoS attribute values into a target value, and finally the best service will be selected by sorting function, that is to say, the service with maximum QoS value can be invoked by users. However, in these methods, the users’ concurrent requests are rarely taken into consideration and the selection mechanisms are not applicable to concurrent requests. Therefore, the method of Top-k service selection based on QoS is proposed. The idea of Top-k service selection method is: Selecting k optimal services from the candidate services with the same functionality, and then randomly assigning to the concurrent requesters. In this way, it is unnecessary to use specific sorting function and it reduces the service response time efficiently.This thesis focuses on the normalization for the qualities of web services, and the Top-k service selection based on QoS. The main research contents and contributions of the thesis are as follows:(1) The existing QoS computation models are analyzed and the normalization method of QoS is improved. During the normalization of QoS attributes, the maximum normalized value of each attribute is set as a variable instead of a constant. It avoids the case that the identical QoS value may be obtained by different candidate services, it is more appropriate for dynamic service environment.(2) The score of each service is computed according to the domination relationship among services. This algorithm introduces two matrices Q and V to represent comparison information and comparison results for each pair of services, respectively. It avoids the duplicated comparison between two services, consequently rapidly computes the service score from the candidate services.(3) A variety of service selection algorithms is analyzed and Top-k web service selection algorithm based on the QoS is proposed. On the basis of aR-tree index, this method decreases the set of candidate services by pruning of depth first searching. By resolving the concurrent requests effectively, Top-k service selection improves the response time that traditional methods cannot reach. As a result, it guarantees the stability to QoS and raises the efficiency of service selection.