面向云计算环境的能耗测量和管理方法

云计算引领了计算机科学的一场重大变革,但与此同时,也不可避免地带来了日益凸显的能源消耗问题,因此,云计算能耗管理成为近几年的研究热点.云计算系统的能耗测量和管理直接关系到云计算的可持续发展,能耗数据不仅关系到能耗模型的建立,而且也是检验云计算资源调度算法的基础.为此,在广泛研究现有能耗测量方法的基础上,归纳总结了当前云计算环境的4种能耗测量方法:基于软件或硬件的直接测量方法、基于能耗模型的估算方法、基于虚拟化技术的能耗测量方法、基于仿真的能耗评估方法,并分析和比较了它们的优势、缺陷和适用环境.在此基础上,指出了云计算能耗管理的未来重要研究趋势:智能主机电源模块、面向不同类型应用的能耗模型、混合任务负载的能耗模型、可动态管理的高效云仿真工具、动态异构分布式集群的能耗管理、面向大数据分析处理和任务调度的节能方法以及新能源供电环境下的节能规划,为云计算节能领域的研究指明了方向.

Energy Consumption Measurement and Management in Cloud Computing Environment

Cloud computing is leading a revolution in computer science, but at the meantime, the large-scale ecosystem of cloud infrastructure brings about the problem of huge energy consumption. Therefore, the energy consumption management has been a research hotspot recent years and to a large extend, the sustainable development of cloud computing has been tightly associated with the techniques of energy consumption measurement and management. Furthermore, the technique of power measurement in cloud environment is the foundation for the building of energy models as well as the evaluation of resources scheduling algorithms. In this paper, based on the survey of a wide range of methods measuring energy consumption of VMs, hosts or the whole system, we summarize four effective ones that are widely applied in cloud system, they are direct measuring techniques based on hardware or software, energy consumption estimate based on energy model, energy consumption measurement under virtualized environment and energy consumption assessment based on simulation technology respectively. We analyze and compare the advantages, drawbacks and best-fit situations of these methods. In addition, we further discuss and point out the trends of future researches on energy management. These trends include smart power supply module, application type oriented energy consumption models, energy consumption model designed for mixed workloads, efficient cloud simulation toolkit for dynamic management, energy management in dynamic heterogeneous distributed cluster, energy preservation with resources scheduling towards tasks processing big data and power provision scheduling with green energy.