基于有色网的动态负载下高性能计算机作业调度系统的性能分析.pdfVIP

基于有色网的动态负载下高性能计算机作业调度系统的性能分析 1,2+ 1 3 李智佳 , 焦莉 , 李焱 1( 中国科学院软件研究所计算机科学国家重点实验室,北京 100190) 2( 中国科学院大学 计算机与控制学院,北京 100049) 3(清华大学 计算机科学与技术系,北京 100084) * Performance analysis for job scheduling systems in HPC Using CPN with Dynamic Workloads 1,2+ 1 3 Li Zhijia , Jiao Li , Li Yan 1(State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing 100190, China) 2(School of Computer and Control Engineering, University of the Chinese Academy of Sciences, Beijing 100049, China) 3(Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China) + Corresponding author: Phn: +86 Fax: +86-10 E-mail: lizj@ Abstract: Performance analysis is an important support technology in the field of computer science. It plays an important role in many stages of life cycle in system, including requirement analysis, design, implement, deploy, optimize and purchase. It can help reveal system bottlenecks and guide system optimization. This paper proposes a performance analysis method for job scheduling systems in high performance computer (HPC) using coloured Petri nets. This method supports dynamic workload, multiple queues and multiple scheduling strategies. As a model method, it has many advantages, such as easy implementation, scalability and low cost. Experimental results show that it is feasible and reliable. Key words: Coloured Petri nets; HPC; job scheduling; performance analysis; simulation ; workload 摘 要: 性能分析是计算机科学领域的一个重要支撑技术，在系统的需求分析、设计、实现、部署、优化、采购等生命周期过程中 发挥重要作用。它有助于揭示系统性能瓶颈，从而指导系统优化。本文提出一种基于有色Petri 网的高性能计算机作业调度系统的性 能分析方法。该方法支持动态作业负载，支持多队列及多种调度策略。作为一种模型方法，该方法具有易实施、易扩展、低成本等 诸多优点。实验表明该方法是可靠的和可行的。 关键词: 有色Petri 网;高性能计算机;作业调度;性能分析;仿真; 负载 1 引言