基于智能代理的网格服务调度研究

Research on grid-service scheduling based on intelligent Agent

为有效解决网格计算环境中的负载平衡和容错问题,提出了面向传统和服务执行方式的分布式调度模型。在该模型中,采用冗余和智能代理技术,以解决网格服务级的负载平衡和容错,同时兼顾了任务级的运行情况。通过引入并定义信誉度的概念,简洁地表征网格资源和网格服务的可用性,同时将其作为调度决策和代理实施的重要工作参数。系统利用环境监测代理检测网格环境,设计平衡代理以辅助调度器来完成动态负载调整和容错。各代理之间利用代理通信语言来进行信息交换。进行相宜的模拟试验,结果表明此模型及其辅助机制能较好地适应网格环境。

To deal with the load balancing and fault tolerance in grid computing environment, a distributed scheduling model oriented to tradition and grid service execution mechanisms was proposed. In this model, redundancy and intelligent Agent were applied to enable service-level load balancing and fault tolerance, whilst operation process of the task-level case was considered. To improve the scheduling efficiency, a new definition, degree of credit, was introduced to concisely denote availability of grid resources and grid services, and it was also an important parameter in decision-making and Agent implementation. Context Agents were responsible for sensing the context of the grid, and balance Agents were enforced to automatically serve as the scheduler for dynamic load adjustment and fault tolerance. Among Agents, information exchanges were conducted via an Agent communication language. Corresponding simulation tests were implemented and their results indicated that this model and its supplementary mechanisms were suitable to grid environments.