超算中心典型科研应用特征统计与分析  

Statistics and Analysis of Typical Scientific Research Application Characteristics of Supercomputing Centers

作　　者：梁润秋 沈瑜[2] Naji Alhusaini 何卓骋 李京[1,2] LIANG Run-qiu;SHEN Yu;Naji Alhusaini;HE Zhuo-cheng;LI Jing(College of Computer Science and Technology,University of Science and Technology of China,Hefei 230026,China;Supercomputing Center,University of Science and Technology of China,Hefei 230026,China)

机构地区：[1]中国科学技术大学计算机科学与技术学院,合肥230026 [2]中国科学技术大学超级计算中心,合肥230026

出　　处：《小型微型计算机系统》2022年第6期1121-1127,共7页Journal of Chinese Computer Systems

基　　金：国家高性能计算环境服务化机制与支撑体系研究项目(二期)(2018YFB0204005)资助.

摘　　要：超算中心已经成为众多科研机构的重要科研基础设施,对基于MPI的高性能计算应用进行数据建模和分析对优化网络流量、资源调度等许多方面十分重要.已有的研究工作大都集中在超算中心大规模MPI整体使用情况,或小规模的某个特定应用的研究,缺乏对科研行业超算中心的研究.本文通过采集和跟踪中国科学技术大学超算中心最典型的两个应用的流量信息:小规模并行的VASP应用和大规模并行的OpenFOAM应用,对MPI各个函数调用的流量数据和数据包传输进行了统计分析并发现其特点和性质,得出了一些新结论,这些结论可以用于指导超算中心网络设计和优化.Supercomputing centers performance optimization have become an important scientific research infrastructure for many scientific research institutions.To this end,Data modeling and analysis of MPI-based high-performance computing applications are very important for optimizing network traffic,resource scheduling and many other aspects.Most of the existing research work focuses on the overall use of large-scale MPI in supercomputing centers,or the research on a small-scale specific application.However,there is a lack of research on supercomputing centers in the scientific research industry.Thus,this paper considers the traffic characteristics and properties of an education organization,University of Science and Technology of China(USTC).In this paper,we choose the two most typical applications of the Supercomputing Center of USTC,the small-scale parallel VASP application and the large-scale parallel OpenFOAM application.First,The traffic information of the theses two applications has been collected,as well as the traffic data and data packet transmission of each function call of MPI are carried out.Secondly,we conducted statistical analysis on the collected data,and drew its characteristics and nature,and reached some new conclusions.These conclusions can be used to guide the operators and developers for more suitable optimization of the supercomputing center network.

关 键 词：超算中心 消息传递接口 网络流量 特征 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]