基于OpenMP的多机集群并行计算系统架构设计

Architecture Design of Multi Machine Cluster Parallel Computing System Based on OpenMP

随着计算机行业的飞速发展,多核处理器已经广泛应用于各种计算机系统,然而传统的单机串行程序依然使用单核计算,无法充分利用多核处理器的性能,更无法使用多机联合计算,在当今大数据的时代满足不了大规模计算的需求。本文针对天地一体化信息网络对大数据处理的需求,基于OpenMP(共享存储并行编程)技术,提出了多核并行程序设计方法,并在此基础上设计了一种多机集群并行计算系统,旨在充分利用计算机性能,提高数据处理和计算的效率,并通过实验分析证明了并行计算系统方案的可行性。

With the rapid development of computer industry,multi-core processors are widely applied to various computer systems.However,the computing used in traditional single-computer serial programs is still the single-core computing,leading to an inadequate utilization of the multi-core processor’s capacity,and even preventing the multi-computer joint computing,which is unable to satisfy the demand for large-scale computing.To meet the requirement for big data processing in the Space-Ground Integrated Information Network,a multi-core parallel programming method based on the OpenMP technology is proposed in this paper.On the basis of the proposed method,a multi machine cluster parallel computing system is designed to make full use of computer capacity and to improve the efficiency of the data processing and the computing.Finally,feasibility of the proposed system is proved by experiments.