刀片式服务器集群性能测试与优化技术研究

发布时间：2018-01-05 15:13

本文关键词：刀片式服务器集群性能测试与优化技术研究　出处：《北京邮电大学》2012年硕士论文　论文类型：学位论文

【摘要】：本文描述的是针对刀片式服务器集群的性能测试与优化技术研究。本文介绍了刀片式服务器集群的构成和工作原理,刀片式服务器集群性能测试的意义、方法和工具,论述了刀片式服务器集群性能测试过程和不同方面的优化方法与过程。研究方法是在了解集群理论知识和计算机集群系统发展动态的基础上,通过大型集群系统实测技术的实践研究,并加以总结和改进。最后通过各种优化措施,提高了刀片式服务器集群性能。与此同时本文还分析了计算机集群系统关键技术的特点。例如：集群集中管理,Infiniband技术,高可用技术,负载均衡技术等。在总结高性能集群,高可用集群,负载均衡集群等不同的集群应用类型的基础上,归纳了计算机集群系统并行计算的常用系统架构：SMP, MPP,集群系统架构等。同时还分析了不同的集群并行环境：MPI, PVM环境,以及其他环境的性能。并对以上各种环境进行了实际操作研究。重点在于集群的测试方法和优化方法的分析。对于Linpack的测试方法和优点进行了长时间的、深入的研究,从而达到提高集群性能的目的,这部分工作是论文主要描述的部分。论文研究工作的主要成果是：分析和归纳了计算机集群系统的组成,并行环境和测试技术；对计算机集群的关键技术进行了深入的分析；对各种测试软件和其配置方法加以深入研究,做到了更好地测试集群性能,并优化集群性能的技术,提高了集群的性能。
[Abstract]:This paper describes the research on performance testing and optimization technology of blade server cluster. This paper introduces the composition and working principle of blade server cluster and the significance of performance testing of blade server cluster. Methods and tools, this paper discusses the performance testing process of blade server cluster and the optimization methods and processes in different aspects. The research method is based on the knowledge of cluster theory and the development of computer cluster system. Through the practical research on the measurement technology of large-scale cluster system, it is summarized and improved. Finally, the performance of the blade server cluster is improved by various optimization measures. At the same time, this paper also analyzes the characteristics of the key technologies of computer cluster system, such as cluster centralized management, Infiniband technology, high availability technology. On the basis of summarizing different cluster application types such as high performance cluster, high availability cluster, load balancing cluster and so on, this paper summarizes the common system architecture of parallel computing in computer cluster system: SMP. MPP, cluster system architecture and so on. At the same time, we analyze different cluster parallel environment: MPI, PVM environment. And the performance of other environments. The practical operation of the above environments is also studied. The emphasis is on the analysis of the test methods and optimization methods of the cluster. The test methods and advantages of the Linpack are given for a long time. Between. In order to achieve the purpose of improving cluster performance, this part of the work is the main part of the paper. The main achievements of this paper are as follows: the composition of computer cluster system, parallel environment and testing technology are analyzed and summarized. The key technology of computer cluster is analyzed deeply. Various test software and its configuration methods are deeply studied, and the performance of the cluster is improved by testing the performance of the cluster better and optimizing the performance of the cluster.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.5

【参考文献】