特定高性能计算集群节能降耗管理研究与应用

发布时间：2018-01-07 04:36

本文关键词：特定高性能计算集群节能降耗管理研究与应用　出处：《兰州大学》2013年硕士论文　论文类型：学位论文

【摘要】：党的十八大报告中,把“生态文明建设”独立成篇,提出了“五位一体”的社会主义建设总布局,促进资源节约,提高利用效率和效益,加强节能降耗,建设资源节约型、环境友好型社会的理念已深入人心。为了提高甘肃省超级计算中心高性能计算集群的服务水平与核心竞争力,笔者在计算集群节点稳定运行的基础上,以实际工作需要为出发点,收集计算集群及相关辅助服务设备的原始数据,监测、分析、模拟、推理这些数据之间的关联情况,从集群虚拟化配置、软件(程序)优化到改善集群的机房运行环境以及强化制度管理等四个方面入手,层层跟进,建立分析问题的模型,实践并优化解决问题的技术方法,得出解决特定高性能计算集群节能降耗切实可行的最佳方案。在实际研究应用中,笔者在有针对性地对集群服务器虚拟化配置和集群系统作业调度(程序控制)做了一些优化后,发现集群在节能降耗方面并没有很明显的变化。考虑到兰州地区季节温差大的特点,遂决定引进新风系统作为节能降耗的主要手段,利用机房各辅助服务设备智能联动,尽量控制集群系统及各种辅助设备引起的机房能耗指数变化,提高集群和机房整体利用率,降低设备能耗,更好地发挥高性能计算集群效能。强化机房制度管理,为用户提供优质高效的服务,达到“资源节约型、环境友好型”绿色高性能计算集群的建设与应用目的,为单位节约运行成本,为社会做出有益贡献。
[Abstract]:In the report of the 18 National Congress of the Communist Party of China, the "Construction of Ecological Civilization" has been put forward as an independent part, and the overall layout of socialist construction of "five in one" has been put forward to promote the conservation of resources, improve the efficiency and efficiency of utilization, and strengthen energy conservation and consumption reduction. The idea of building a resource-conserving and environmental-friendly society is deeply rooted in people's hearts. In order to improve the service level and core competitiveness of the high-performance computing cluster in Gansu supercomputing center, the author takes the practical work as the starting point on the basis of the stable operation of the computing cluster node. Collect raw data for computing clusters and related assistive service devices, monitor, analyze, simulate, and infer the correlation between these data from cluster virtualization configuration. Software (program) optimization to improve the cluster of computer room operating environment and strengthen the four aspects of the system management, layer by layer follow up, set up an analysis of the problem model, practice and optimize the technical methods to solve the problem. Finally, the best scheme to solve the problem of energy saving and consumption reduction in special high performance computing cluster is obtained. In the actual research and application, the author has made some optimization to the cluster server virtualization configuration and the cluster system job scheduling (program control). It is found that there is no obvious change in energy saving and consumption reduction in the cluster. Considering the characteristics of large seasonal temperature difference in Lanzhou area, it is decided to introduce the fresh air system as the main means of energy saving and consumption reduction. Make use of the intelligent linkage of each auxiliary service equipment in computer room, try to control the change of energy consumption index caused by cluster system and various auxiliary equipments, improve the overall utilization ratio of cluster and computer room, and reduce the energy consumption of equipment. Improve the efficiency of high-performance computing cluster, strengthen the management of computer room system, provide users with high quality and efficient service, and achieve the purpose of "resource saving, environment friendly" green high performance computing cluster construction and application. Save running cost for the unit, make beneficial contribution for the society.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP308

【参考文献】