高校中小型高性能集群系统的建设及管理实践  

Construction and Management Practice of Small and Medium-sized High-performance Computing Cluster System in Colleges and Universities

作　　者：冯伟[1] 姜远飞[1] 孙晶[2] 姚震[3] 刘爱华 Feng Wei;Jiang Yuanfei;Sun Jing;Yao Zhen;Liu Aihua(Institute of Atomic and Molecular Physics,Jilin University,Changchun,130012,China;School of Mathematics,Jilin University,Changchun,130012,China;State Key Laboratory of Superhard Materials,Jilin University,Changchun,130012,China)

机构地区：[1]吉林大学原子与分子物理研究所,吉林长春130012 [2]吉林大学数学学院,吉林长春130012 [3]吉林大学超硬材料国家重点实验室,吉林长春130012

出　　处：《中国现代教育装备》2024年第17期16-19,共4页China Modern Educational Equipment

基　　金：科技部国家重点研发计划(政府间国际科技创新合作重点专项)“基于阿秒脉冲的纳米结构与富勒烯团簇超快动力学理论研究”(编号:2022YFE0134200);吉林大学实验技术项目“高性能计算平台的整合及远程监控管理系统的开发实施”(编号:2023-SYXM2023b023);吉林大学超硬材料国家重点实验室开放课题“超硬材料理论研究平台计算性能及管理效率提升的研究”(编号:202425)。

摘　　要：理论模拟实验室是高校各级科研单位开展教学和科研工作的重要支撑平台。如何建设可持续发展且具有高性价比的高性能计算集群系统,并实现集群系统的高可用性和安全使用是各级科研单位集群系统建设和管理的重要问题,也是理论模拟实验室在“双一流”建设中所要面对的重要任务之一。基于吉林大学原子与分子物理研究所共享集群系统的建设、管理和维护实践,提出集群软件等公用资源共享使用、计算资源统一规划、分级管理,并结合自行开发的集群监控管理系统,利用手机端进行集群状态监控,发现异常即时处理,从而提高集群系统管理效率和确保集群系统稳定运行,为学科发展提供重要支撑。The theoretical simulation laboratory is an important support platform for scientific research and related teaching work carried out by various levels of scientific research units in universities.How to build a sustainable and cost-effective high-performance computing cluster system,and achieve high availability and safe use of the cluster system is an important issue for the construction and management of clusters in scientific research units at all levels,and is also one of the important tasks faced by theoretical simulation laboratories in the "Double First-class" construction project.Based on the construction,management,and maintenance practice of the shared cluster in the Institute of Atomic and Molecular Physics of Jilin University,this paper proposes a maintenance and management approach for the shared use of public resources such as cluster software,unified planning,and hierarchical management of computing resources,and combined with a self-developed cluster monitoring and management system,cluster status monitoring can be carried out on the mobile end.Any abnormalities found can be immediately handled,thereby improving the management efficiency of the cluster system,ensuring its stable operation,and providing important support for the development of the discipline.

关 键 词：理论模拟 高性能计算集群 实验室管理 分布式拒绝服务 监控管理系统 

分 类 号：TP308[自动化与计算机技术—计算机系统结构] G482[自动化与计算机技术—计算机科学与技术]