计算机集成制造系统 ›› 2018, Vol. 24 ›› Issue (第7): 1816-1821.DOI: 10.13196/j.cims.2018.07.023

• 当期目次 • 上一篇    下一篇

WRF与NPB科学计算程序在云平台上的性能研究

张心言1,任开军1,2,赵成武1,邓科峰1,2   

  1. 1.国防科技大学计算机学院
    2.国防科技大学海洋科学与工程研究院
  • 出版日期:2018-07-31 发布日期:2018-07-31
  • 基金资助:
    国家自然科学基金面上资助项目(61572510)。

Performance analysis of WRF and NPB scientific computing programs on scientific cloud platform

  • Online:2018-07-31 Published:2018-07-31
  • Supported by:
    Project supported by the National Natural Science Foundation,China(No.61572510).

摘要: 为了探究IaaS云平台在运行科学计算程序方面存在的优势与不足,采用虚拟化技术基于高性能计算物理集群创建了NUDT_SCICloud科学云平台,并选择通用NPB标准测试集和包括若干数据与计算流程的大规模气象数值模拟软件WRF模式对该平台进行了全面性能测试,最后将测试结果与物理集群进行了分析对比。实验结果表明,NUDT_SCICloud云平台运行大规模科学计算程序具有可行性,但与物理计算集群相比性能有一定的差距。

关键词: 云计算, 高性能计算, 科学云, 天气预报模式

Abstract: To explore the advantages and disadvantages of IaaS cloud platform in running scientific computing programs,NUDT_SCICloud scientific cloud platform was constructed based on High Performance Computing(HPC)physical cluster with the help of virtualization technology.A comprehensive performance test of the platform was carried out with both NAS Parallel Benchmark (NPB) standard test set and large-scale meteorological numerical simulation software Weather Research and Forecast Model (WRF) which included a large number of data processing workflow.Furthermore,the test results were compared with those of the physical cluster and were analyzed,which concluded that it was feasible to run large-scale scientific computing programs on NUDT_SCICloud cloud platform,yet a certain performance degradation exits compared with the physical cluster.

Key words: cloud computing, high performance computing, scientific cloud, weather research and forecast model

中图分类号: