基于自适应微粒群算法的网格工作流调度  

A GRID WORKFLOW SCHEDULING BASED ON SELF-ADAPTIVE PARTICLE SWARM OPTIMIZATION

作　　者：于明远[1] 胡亚红[2] 王子仁[2] 

机构地区：[1]华中科技大学数字化工程与仿真中心,湖北武汉430074 [2]浙江工业大学信息学院,浙江杭州310032

出　　处：《计算机应用与软件》2008年第8期19-21,44,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(60673177);浙江省自然科学基金ZJNSF(Y105109);浙江省教育厅科研项目(20070284)

摘　　要：服务网格是一个通过组合网格服务为用户提供强有力的各种服务的系统,其中网格服务遵循OGSA的标准。网格服务工作流调度的关键在于如何在应用程序运行过程中能动态地根据当前系统中基本服务的情况,组合出满足用户需要的服务。提出了一种自适应微粒群优化算法用于服务感知的Web服务选择,其中引入了一个特殊的速度变异操作来增强空间搜索的有效性,并融合了遗传算法杂交与变异。它不仅能很好地满足组合服务的需求,而且能更有效地进行全局搜索。仿真试验显示对于具有全局Qos约束条件的Web服务选择在执行效率上自适应微粒群优化算法明显优于其它混合遗传算法(如种群多样性控制遗传算法)。Grid Services are Web services following the OGSA （Open Grid Service Architecture） specification, and the Service Grid is a service system providing more powerful services to customers by integrating Grid Services. Key of Grid service-workflow scheduling is that how to make selection of services to better fulfill customer＇s expectations by dynamic combination of various QoS during execution of applications. A heuristic algorithm, self-adaptive particle swarm optimization algorithm （SAPSOA）, is presented for QoS-aware Web services selection. In this paper it uses a special mutation operator to make particles explore the search space more efficiently, and supports ideal of genetic algorithm with crossover and mutation. It can not only get more excellent composite service plan, but also effectively seek the global excellent result. The simulation results on web services selection with global QoS constraints have shown that SAPSOA performs well than some other genetic algorithms（ i.e. population diversity controlled genetic algorithm）.

关 键 词：网格工作流 服务质量感知 自适应微粒群算法 混合遗传算法 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论] TP273.2[自动化与计算机技术—计算机科学与技术]