摘要
使用绿色能源已成为解决数据中心能耗问题的一种有效方式。为了降低绿色能源变化幅度大的特点带来的影响,通常将可延迟作业放入等待队列,将相应空闲服务器置为休眠状态,降低系统能耗,在新能源可用的时候执行作业。当新作业执行时,需要重新开启休眠状态服务器来保证数据可用性。数据放置与作业执行时间的不统一,会导致频繁开启休眠服务器,带来能源浪费。针对绿色数据中心提出一种数据调度策略,根据数据处理型框架中等待队列作业调度次序,通过将未来一段时间内需要被读取的数据块提前复制在活跃服务器上,降低休眠状态服务器开启的次数,从而降低总体能耗。实验模拟结果显示,该算法可平均减少43%的休眠状态服务器重复开启次数。
Using renewable energy in data center is an environment-friendly way to solve the problem of high energy consumption of data center. Since renewable energy is variable, delaying the jobs which has no strict deadline i a widely used strategy to maximize the usage of renewable energy. Meanwhile, turning the idle servers off can further reduce energy consumption. If the data required by the jobs to be processed are unavailable, some servers in sleep state need to be reactivated to guarantee that the data required by the jobs are available. Such operation may lead to energy waste due to the frequent reactive processes. An effective data management algorithm was proposed, which copied the data required by the jobs in waiting queue to active servers in advance. By doing so, the times that the sleep servers were reactivated could be reduced. Simulation results show that the times can be reduced by 43% on average.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2016年第3期592-599,共8页
Journal of System Simulation
基金
国家自然科学基金(61370028
91218302
61321491)
江苏省自然科学基金(BK2011191)
江苏省科技支撑计划(BE2013116)
中央高校基本科研业务费专项资金(20620140514)
国家电网公司科技项目
关键词
绿色数据中心
数据处理型框架
能耗
新能源
数据管理
green data center
data processing framework
energy
renewable energy
data management
