曙光高性能计算机在数值预报模式中的应用  

Application of HPC in Numerical Prediction Model

作　　者：王俊超[1] 彭涛[1] 冯光柳[1] 

机构地区：[1]中国气象局武汉暴雨研究所暴雨监测预警湖北省重点实验室,湖北武汉430074

出　　处：《计算机技术与发展》2014年第10期178-181,185,共5页Computer Technology and Development

基　　金：国家自然科学基金资助项目(41205086);公益性行业(气象)专项(GYHY201206028;GYHY201306056);暴雨监测预警湖北省重点实验室暴雨研究开放基金(IHR201305)

摘　　要：文中首先介绍了中国气象局武汉暴雨研究所高性能计算机应用现状和目前的模式业务系统,针对气象预报模式精细化对计算能力的更高需求,中国气象局武汉暴雨研究所采用曙光高性能计算机集群对原有集群系统进行升级,升级后的计算节点CPU可提供11.40 TFlops的双精浮点计算能力;其次,讨论了升级后的高性能计算机几个关键技术的现状并对未来进行展望;最后以WRF模式为例,对升级后的高性能计算机的性能进行了分析,得到了较好的加速比。结果表明:新升级的集群系统将大大节省区域高分辨数值预报模式运算时间,有助于提高科研成果的转化效率。In this paper,firstly introduce the application of HPC in WHIHR and the current mode of business systems. For the better re-quirements of the weather forecasting model fine to computing ability,the institute of heavy rain of China Meteorological Administration uses the dawn of high-performance computer clusters to upgrade the original cluster system. The upgraded computing node can provide the 11. 40 TFlops double-precision floating-point computing power. Secondly,discuss the present status and the prospect of key technol-ogy of advanced HPC. Finally,taking WRF model as example,analyze the performance of HPC and have a good speedup. The result shows that the new upgraded cluster system would greatly reduce the computation time of regional high-resolution numerical weather pre-diction model. Also it could improve the conversion efficiency of scientific research.

关 键 词：高性能计算机 数值预报模式 并行计算 加速比 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]