基于概念关联程度的网格服务组合方法

Conceptual correlation-based method for grid service composition

为克服自动服务组合在网格环境下性能低下的弱点,提出了基于领域本体的两阶段法(two phases method,TPM)。首先根据聚类规则将同类网格服务归并为一个抽象服务类,针对服务类进行组合;然后选择具体的网格服务替换第一阶段生成的组合方案中相应的服务类。TPM定义了基于领域本体的服务类关联程度,并根据关联关系构建服务组合图,将服务组合问题转化为最短路径问题,给出了基于Dijkstra贪婪搜索的组合算法及服务选择策略。通过实验与同类方法在计算效率、服务满意度和服务质量上进行了对比,实验结果证明TPM规模扩展性良好,并能提供一定的QoS保障。

The performance of automatic service composition in the grid is improved by a domain ontology based two-phase method （TPM）. The method first clusters functional similar grid services into one abstract service class using cluster rules, and the service classes are composited based on their conceptual correlations. Then the system replaces the composite service classes by selected grid services. With the conceptual correlations between service classes defined using domain ontology, the composition of service classes is transformed to a ＂shortest path problem＂ on the service composition graph constructed based on the correlations. A composition algorithm completes the process using I）ijkstra search with a simple service selection strategy. Finally, comparisons of the efficiency, satisfaction, and QoS shows that the method is more efficient and scalable than other methods and provides better QoS.