Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (5): 33-37.

Previous Articles     Next Articles

Relative sustained performance metric model for high performance computing system

XIAO Huadong, SUN Jing, WEI Min, LI Juan, SHEN Yu   

  1. High Performance Computing Division, National Meteorological Information Center, Beijing 100081, China
  • Online:2015-03-01 Published:2015-04-08

高性能计算机系统相对持续性能度量模型

肖华东,孙  婧,魏  敏,李  娟,沈  瑜   

  1. 国家气象信息中心 高性能计算室,北京 100081

Abstract: The sustained performance of a High Performance Computing System (HPCS) is an important metric to reflect the performance of HPCS achieved in the real application of field. The general performance evaluation methods of HPCS are briefly introduced. Combined with application benchmarks, the metric model of relative sustained performance is proposed. The experiment is conducted based on the benchmark of meteorological applications. The result indicates that applying the model can make an obvious distinction of the performance of the HPCS from different vendors, which provides some referenced value for the choice of the appropriate HPCS. The scalability of meteorological applications is also reflected to some extent through the model.

Key words: performance evaluation, benchmark, relative sustained performance

摘要: 高性能计算机系统的持续性能是反映实际领域应用中高性能计算机系统性能强弱的重要度量标准。简单介绍了高性能计算机系统常用的性能评价方法,结合应用基准程序集,提出了相对持续性能的度量模型。实验基于高性能计算气象应用评测,结果表明应用相对持续性能度量模型可区分不同厂商的高性能计算机系统的性能强弱,为高性能计算机系统的选择提供参考,并在一定程度上反映了气象应用本身的可扩展性。

关键词: 性能评价, 基准程序集, 相对持续性能