摘要
数据中心的高能耗是一个亟待解决的问题.尤其是随着云计算的发展,更多的资源集中到云端.构建绿色数据中心、实现节能减排成为了近年来业界关注的热点.数据中心的能耗主要由计算能耗和制冷能耗两部分组成.数据中心的热量管理主要从减少制冷能耗的角度出发,为实现绿色计算提供了新的思路.该文从绿色数据中心的状态监控、热量建模、热量管理策略以及热量管理评价4个方面综述了近年来数据中心热量管理方面的研究工作.该文提出了绿色数据中心热量管理的总体架构,总结了其分布式监控系统的一般框架;对现有的热量管理方法按面向单节点/面向多节点进行分类,并且从复杂度、灵活度、实施效果等多方面进行了比较,分析了各种方法的优势和局限性.文中提出了数据中心全局能耗评价、制冷系统效率评价、热量及温度评价的分类方法,对现有的评价方法进行总结.最后论文列出了未来需要进一步研究的十个方向.
The high energy consumption of data center is a serious problem to be solved.Especially as the development of cloud computing,more resources are centralized to the clouds.Constructing green data centers and achieving power cost and carbon footprint reduction became research hotspots in recent years.Energy consumption of data centers consists of computational energy and cooling energy.Thermal management for data centers mainly target reducing cooling energy and provide new solutions to achieve green computing.This paper presents a review of the recent research work of thermal management from the perspectives of status monitoring,thermal modeling,thermal management policies and thermal management evaluations.We propose the overall architecture of thermal management of green data center,and summarize its general framework of distributed monitoring system. We classify the existing thermal management policies into two classes：single-node case and multiple-node case,compare the complexity,flexibility and effectiveness of the existing policies,and analyze their advantages and limitations.This paper summarizes the existing evaluating approaches proposing a new taxonomy of three categories：global energy consumption evaluation,refrigerating efficiency evaluation,and thermal/temperature evaluation.Finally,ten possible research directions are suggested for future research in this field.
出处
《计算机学报》
EI
CSCD
北大核心
2015年第10期1976-1996,共21页
Chinese Journal of Computers
基金
国家"八六三"高技术研究发展计划重大项目基金(2011AA01A207)
国家自然科学基金(61272128)资助~~
关键词
绿色计算
绿色数据中心
热量管理
能耗管理
制冷
云计算
green computing
green data center
thermal management
energy management
refrigeration
cloud computing
