| Summary: | 碩士 === 國立成功大學 === 資訊工程學系 === 105 === Quality of service (QoS)-aware service composition aims to select an optimal execution plan to maximize the quality values of the resulting composite service while satisfying user-specified constraints. Existing methods on QoS-aware service composition mostly orchestrate multiple services to meet a single user request at a time and neglect serving multiple user requests simultaneously. In the case of multiple user requests at the same time, users will compete for the service with best quality, which cause conflict and overload in QoS-aware service composition. To overcome this shortcoming, we first model the problem as a Time-Constrained Service Composition (TCSC) problem aiming to determine composite services with the maximum number of success user requests. A Multi-User Service Composition System (MUSCS) is then proposed to find feasible solutions, which contains six main components, including User Manager, Request Classifier, Service Controller, Service Classifier, Abstract Service List (ASL) Composer and ASL monitor. Among these components, the ASL Composer plays a major role to deploy candidate services for multiple user requests in a reasonable time. We therefore develop three algorithms for the ASL Composer, namely Time-Constrained Incremental On-Demand Algorithm (TCIDA), User Oriented Incremental On-Demand Algorithm (UOIDA), and Greedy Multiple Matching (GMM). By dynamically updating the similarity between service quality and user quality requirement, the proposed algorithms are able to release the overloaded service and decrease the conflict conditions. The evaluation results demonstrate the effectiveness of our algorithms in term of the number of success user requests.
|
|---|
