多集群并行作业的性能监测及分析

Performance Monitoring and Analysis on Multi-cluster Parallel Jobs

介绍一种多集群计算模式,在分析了多集群系统结构灵活、具有可重组性等特点的基础上,研究适用于该模式的并行作业性能监测分析方法与技术,设计和实现了一个并行作业性能监测分析工具。它采用动态性能分析方法,遵循分布式软件设计架构,具有高内聚、低耦合的模块组织结构,运行验证表明其能够在多集群计算模式下有效工作。

A multi-cluster computing model is introduced. In the analysis on the basis of the multi-cluster system features, which include flexible architecture and reconfigurability, the methods and technologies of performance monitoring and analysis, which is applicable to this model, are researched. A performance monitoring and analysis tool of parallel jobs is designed and implemented. In this tool, dynamic performance analysis method is used, distributed software design framework is followed, a high cohesion and low coupling structure is designed. Operation results show it can work effectively in the multi-cluster computing model.