集群的监控和管理工具多种多样,设计方法也有很多,本文综合国内外多种集群系统 的性能检测工具设计方法的优点,结合面向对象设计思想,主要对Infiniband 网络性能监 控和并行程序监控进行改进,在Redhat Linux 系统下使用C 语言和QT 实现可视化的图形监控和管理工具。 关键词: InfiniBand 集群 系统监控 网络监控 libp_cap 【ABSTRACT】Monitoring and management tools of clusters are diverse, and their designs are also various. This article unites many good designs of cluster performance monitoring system with object-oriented programming. It is improved on InfiniBand network performance monitoring and parallel program monitoring, what is developed to be a graphical monitoring and management tool with C and QT in Redhat Linux system. 【KEYWORDS】InfiniBand, cluster, System monitoring, network monitoring, libp_cap