WRF模式在LINUX集群系统的并行计算与评测  

Parallel Computing and Evaluation of WRF Model on LINUX Cluster System

作　　者：李俊醅[1] 庄子波[1] 

机构地区：[1]中国民航大学飞行技术学院,天津300300

出　　处：《计算机技术与发展》2012年第7期5-8,共4页Computer Technology and Development

基　　金：国家自然科学基金(41075013);中央高校基金(ZXH2010D021;ZXH2009B001)

摘　　要：WRF气象模式是高性能的并行计算,一般需要大型机进行并行计算和处理,文中将WRF系统构建在32核LINUX集群系统上,进行并行计算试验并确定出了该系统的合理的并行计算方式。选取了WRF个例分别对运算时间、加速比和并行效率做了性能测试。结果表明:处理器个数逐步增加至16时,WRF系统运算的总时间随个数单调减小,从16增加到26,系统运算的总时间变化总体趋势在减小,但变化不明显,存在一定的起伏,从26到32,总体上有增加的趋势。实验构建的集群系统具有较好的并行计算效率和加速比,当处理器个数为26时可以达到最大的加速比12.5,且并行效率达到48%,能够达到较好的并行运算效果。WRF meteorological model is a high-performance parallel computing, generally require mainframe computing and parallel pro- cessing. This system will be built on the 32 WRF Core Xeon cluster system. One WRF patients was selected to evaluate the system per- formance of the total computation time ,parallel efficiency and acceleration rate. The results showed that： with the gradual increase in the number of processors to 16 ,the total time decreases monotonically with the number of processors. From 16 to 26 ,the overall trend of the total time is decreasing,but did not change significantly. From 26 to 32,the overall trend of the total time is increasing. The experimental cluster system has good parallel efficiency and acceleration rate. When the processor is 26, the better parallel performance can be a- chieved,maximum acceleration rate is 12.5 ,and parallel efficiency is 48%.

关 键 词：并行计算 加速比 集群系统 WRF模式 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]