集群负载平衡的性能评测方法

Performance benchmark methodology for load balancing cluster

随着单一系统映像(SSI)集群的发展,其提供的易于使用的高性能和高可用的计算环境,对用户越来越具有吸引力。但是,对于单一系统映像集群,现有的基准测试程序在测试系统性能时将负载静态地平均到各个处理器,而没有考虑系统的动态负载平衡特性。基于此,提出一种针对负载平衡集群的性能评测方法,采用非平衡树搜索算法提供动态负载,通过负载单元的运行时间评测系统在真实的负载平衡环境下的性能。

Development of single system image （SSI） cluster provides a easy to use, high performance and high available computing ＂environment, which has become increasingly attractive to users. However, existing benchmark allocates the tasks staticly to the processors when testing the performance of the SSI cluster, without consideration of the dynamic load balancing on the system. In view of the above, a performance benchmark methodology is presented for a load balancing SSI cluster, measuring load unit run time and evaluating system performance on a substantial load balancing condition with umbalanee tree search （UTS） to provide load to system.