基于Hadoop分布式计算平台的磁流体动力学模型仿真研究

在Hadoop分布式云计算平台上进行科学计算仿真,具有节省软硬件投资、缩短模拟时间等研究意义。针对需要高计算能力的磁流体动力学(MHD)仿真问题,设计了一种基于Hadoop分布式计算平台的MHD仿真器。首先,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。然后,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。

Simulation of Magneto Hydro Dynamic Model Based on Hadoop Distributed Computing Platform

Running scientific computing simulation on the Hadoop distributed cloud computing platform has many research meanings such as saving hardware and software investment and shortening simulation time. For the issues that the magneto hydro dynamic (MHD) simulation which need high computing power, a MHD simulation based on Hadoop distributed computing platform is proposed. First of all, the two kind of distributed parallel computing models: Spark and HAMA are integrated into the Hadoop ecosystem, to support memory computing and synchronous parallel computing. Then, the Hadoop is built as a Master-Slave peer-to-peer structure, to solve global synchronization and local synchronization. Finally, the finite volume method and Riemann are used to solve the problem of the MHD equation. The experimental results show that the simulator can accurately simulate MHD, and greatly shorten the simulation time.