中文摘要：

针对磁流体动力学（MHD）仿真需要高计算能力的问题,设计了-种基于Hadoop分布式计算平台的MHD仿真器.将Spark和Hama两种分布式并行计算模型整合到Hadoop生态系统中,分别用于支持内存计算和整体同步并行计算.将Hadoop集群构建成master-slave对等结构,解决全局同步和局部同步问题.最后,在Hadoop集群上,利用有限体积法和黎曼问题来求解MHD方程.实验结果表明,该仿真器能够精确模拟MHD,同时大大缩短了仿真计算时间.

英文摘要：

For the issues that the magneto hydro dynamic （MHD） simulation need high computing power, this papera MHD simulator based on Hadoop distributed computing platform. First of all , this simulator combined the two kind buted parallel computing models： Spark and Hama into the Hadoop ecosystem, to support memory compparal lel computing. Then , it bui lt the Hadoop as a master-slave peer-to-peer structure , to solve global synchronization and lo-cal synchronization. Finally,it used the fini te volume method and Riemann to solve the problem of the MHD equation. Theexperimental results show that the simulator can accurately simulate MHD , and greatly shorten the simulation