计算机与现代化

• 软件工程 • 上一篇    下一篇

高性能计算集群的Linpack测试及其在大数据中的应用

  

  1. (中国石油大学(北京)地质地球物理综合研究中心,北京 102249)
  • 收稿日期:2015-02-04 出版日期:2015-05-18 发布日期:2015-05-18
  • 作者简介:韩菲(1985-),女,黑龙江桦南人,中国石油大学(北京)地质地球物理综合研究中心博士研究生,研究方向:高性能计算,GPU并行计算; 通信作者:孙赞东(1961-),男,湖北黄冈人,教授,博士生导师,博士,研究方向:人工智能及应用,可视化技术,应用地球物理方法; 苏辉(1990-),男,系统工程师,本科,研究方向:系统安全与维护。
  • 基金资助:

Linpack Test of High Performance Computing Cluster and Its Application in Big Data

  1. (Laboratory for Integration of Geology and Geophysics, China University of Petroleum, Beijing 102249, China)
  • Received:2015-02-04 Online:2015-05-18 Published:2015-05-18

摘要: 高性能计算集群用于高效并行计算,具有很高的性价比和良好的可扩展性,如何测试和评价集群系统性能成为一个关键问题。本文基于6个节点的集群进行Linpack测试,测试不同问题规模、计算节点数、求解矩阵数据分块NB、处理器网格拓扑P×Q、网络通信等重要因素,将单机与集群的计算性能进行对比,测试集群性能,结果表明:该集群的并行计算性能良好,可扩展性强,但硬件通讯能力需进一步改善。应用该集群到实际的地震大数据计算中,该集群的并行计算能力得到了很大的提升。

关键词: Linpack, 高性能计算, 性能测试, 大数据

Abstract: High performance computing cluster for efficient parallel computing is of very high cost performance and good scalability. How to test and evaluate the cluster system performance becomes a key problem. In this paper, we made a Linpack test based on high performance computing cluster of 6 nodes. Different matrix scale N, computational nodes, block size NB of matrix, processor grid topology and network communication were tested. Computing performance to PC and cluster were compared and cluster performance was tested. The results show that: the cluster parallel computing performance is good and is of strong scalability, but the hardware communication ability should be further improved. The cluster is applied into the computing of the actual seismic big data. It is known that the parallel computing performance of the cluster has been improved greatly.

Key words: Linpack, high performance computing, performance test, big data

中图分类号: