基于应用的全局资源优化调度方法  

Application based global resource optimization scheduling method

作　　者：寇大治[1] 沈瑜 唐小勇 KOU Dazhi;SHEN Yu;TANG Xiaoyong(Department of High Performance Computing,Shanghai Supercomputer Center,Shanghai 201203,China;School of Computer Science and Technology,University of Science and Technology of China,Hefei Anhui 230026,China;College of Information Science and Engineering,Hunan University,Changsha Hunan 410082,China)

机构地区：[1]上海超级计算中心高性能计算部,上海201203 [2]中国科学技术大学计算机科学与技术学院,合肥230026 [3]湖南大学信息科学与工程学院,长沙410082

出　　处：《计算机应用》2019年第S02期156-159,共4页journal of Computer Applications

基　　金：国家重点研发计划项目(2018YFB0204004)

摘　　要：在国家高性能计算环境中,为了更好地实现对分布在不同地域超级计算机资源的调度管理,针对计算资源忙闲不均等问题,提出通过研究典型应用作业的运行特征,开发多中心任务的调度系统,以解决国家高性能计算环境统一调度的关键技术问题。首先收集了若干超级计算中心的应用运行历史情况;然后研究了高性能计算系统的历史任务数据,建立应用运行历史数据库;最后将用户应用对资源的需求和典型应用的资源使用特征分析相结合,建立一种可精确描述应用特征的框架。研究了基于多中心应用特征的任务调度方法,开发了基于应用的全局资源优化调度系统,为国家高性能计算环境服务化运营和稳定运行提供了有力的技术支撑,有效地提高了国家高性能计算环境的可靠性、可用性和可维护性。In the national high performance computing environment,the supercomputer resources distributed in different regions are dispatched and managed.In order to avoid the problem of busy and unevenly distributed computing resources,it is necessary to develop a multi-center task scheduling system by studying the running characteristics of typical application operations to solve the unified regulation of national high performance computing environment.Firstly,the application running history of several supercomputing centers were collected,then the historical task data of high performance computing system were studied,the application running history database was established,and finally the user application s demand for resources was combined with the analysis of resource usage characteristics of typical applications.A framework for accurately describing application characteristics was established,and the application characteristics based on multi-center were studied.Task scheduling method and application-based global resource optimization scheduling system were developed,which provide powerful technical support for the service and stable operation of national high-performance computing environment,and effectively improve the reliability,availability and maintainability of national high-performance computing environment.

关 键 词：超级计算 高性能计算系统 历史数据库 应用特征 调度方法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]