高性能集群计算系统的构建  

Establishment of High Performance Cluster Computing System

作　　者：李圣强[1,2] 李闽峰[2] 刘桂平[2] 王斌[2] 吴婷[2] 王浩[2] 

机构地区：[1]中国科学技术大学地球与空间科学学院,安徽合肥230026 [2]中国地震局地震预测研究所,北京100036

出　　处：《地震》2012年第1期144-149,共6页Earthquake

基　　金：中国地震局地震预测研究所基本科研业务费(0207690210)

摘　　要：高性能集群计算系统因其具有强大的运算能力、较高的I/O性能、高性能管理和较强的系统扩展能力而广受关注。本文简要介绍了高性能计算的发展现状和高性能集群计算系统构成,比较详细地描述了系统环境、系统性能及集群管理与应用情况,对高性能计算的新技术进行展望。并以一个实例阐述了集群系统构建等内容。该系统由192个计算节点、4个I/O节点、1个登录节点、1个管理节点、1个Windows集群服务节点、1组盘阵、1组计算与I/O网络和1组管理网络组成。系统计算能力在2009年和2010年中国高性能计算机性能TOP100排行榜分别为第36位和第63位,具有相当的竞争力,能够在近几年内为地震科研提供有力支持。High-performance cluster computing systems （HPCC） is widely accepted because of its powerful computing capability, high I/O performance, and strong performance management system scalability. In this paper, the development and topological composition of high performance computing system are briefly introduced. System environment, performance features, HPC management and its applications of this supercomputational platform are comparably described in detail. As an example, we presented the system constructed in IES consisted of 192 computational nodes, 4 I/O nodes, 1 login node, 1 management node, 1 windows cluster service node, a group of disk array, a group of computing, I/O management and network. The China TOP100 Rank of High Performance Computer of IES released by the Specialty Association of Mathematical ＆ Scientific Software （SAMSS） ,CSIA is listed in 36 and 63 respectively in 2009 and 2010, which proved that it has competitive capacity and can provide powerful support for earthquake scientific research tasks in recent years.

关 键 词：拓朴结构 系统环境 集群管理 高性能计算系统 

分 类 号：P315.6[天文地球—地震学]