基于弹性云计算集群的国际TEAM Problem 7基准问题计算方法

发布时间：2018-03-08 06:16

  本文选题：云计算　切入点：高性能计算　出处：《电工技术学报》2017年08期 　论文类型：期刊论文

【摘要】：针对智能制造所需的大规模计算问题,对比分析不同云计算技术的发展模式,选取适合大规模计算密集型的云架构,部署了9节点弹性计算平台。选取电磁领域的国际TEAM Problem7基准问题为案例,使用基于消息传递机制的开源并行编程模型Open Mpi编写典型涡流场的并行计算程序,实现了计算案例数百万自由度的并行计算,得到了不同节点下不同计算规模的计算时间与速率分布图。研究结果表明,典型电工装备的大规模有限元仿真计算可在云计算平台部署的弹性计算集群中实现并行化。对比于传统高性能计算机的并行计算方法,本文的云计算弹性计算集群具有投入成本低、动态可分配、按需索取的优点,为传统纯粹制造业转型升级智能制造产业发展提供了一种灵活、便捷的新型高性能计算方案。
[Abstract]:Aiming at the problem of large-scale computing for intelligent manufacturing, this paper compares and analyzes the development patterns of different cloud computing technologies, and selects a cloud architecture suitable for large-scale computing intensive. A 9-node elastic computing platform is deployed. The international TEAM Problem7 benchmark problem in electromagnetic field is selected as an example, and the parallel computation program of typical eddy current field is compiled by using the open source parallel programming model Open Mpi based on message passing mechanism. The parallel computation of millions of degrees of freedom is realized, and the calculation time and rate distribution of different computing scales under different nodes are obtained. The results show that, The large-scale finite element simulation of typical electrical equipment can be parallelized in the elastic computing cluster deployed on cloud computing platform. The flexible computing cluster in this paper has the advantages of low input cost, dynamic distribution and on-demand demand. It provides a flexible and convenient new high performance computing scheme for the traditional pure manufacturing industry to transform and upgrade the intelligent manufacturing industry.
【作者单位】： 天津工业大学天津市电工电能新技术重点实验室;
【基金】：国家自然科学基金面上项目(51577132) 国家火炬计划(2015GH611592)资助
【分类号】：TP38
，

本文编号：1582747