云数据中心可靠性感知负载均衡调度算法  

Reliability-Aware Load Balance Scheduling Algorithm for Cloud Data Center

作　　者：敬超[1,2] 宋琪[1] 

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004 [2]上海交通大学电子信息与电气工程学院,上海200240

出　　处：《计算机仿真》2016年第6期208-212,330,共6页Computer Simulation

基　　金：国家自然科学基金(61563012;61540054);广西自然科学基金(2015GXNSFBA139260);桂林理工大学科研启动基金(002401003456)

摘　　要：绿色云数据中心的建设,虽然缓解了数据中心高能耗的问题,但是内部计算节点能耗高,使得计算节点温度升高可靠性差,运行过程中失效无法完成任务。然而在数据中心内由于冷却系统制冷的不均匀,使得不同位置计算节点的温度差异性大。针对上述问题,提出通过热容阻模型将温度与可靠性相互关联建立计算节点的可靠性模型。为了保证节点温度的平衡,利用概率论的离散方差计算理论,计算各节点间的温度偏差,从而为均衡负载提供指导思想,提出了一种面向云数据中心的可靠性感知负载均衡调度算法。通过仿真,将提出算法分别与经典的调度算法以及最新的算法进行了比较,实验结果表明提出的算法提高了太阳能利用率,而且总能耗使用合理保证了节点的可靠性及作业时限。Despite the green data centers are appearing, the energy consumption for data center is still rocket growing. It will result in the poor reliability of computing nodes that substantially failure during the operating process. The cooling system also leads to uneven temperature among different nodes with various positions. This paper firstly leverages thermal resistance- capacity model that associates the temperature and reliability to establish the reliable model for computing nodes. Then the discrete temperature variance based on probability theory is implemented for computing the deviation temperature on various nodes that substantially leads to load balance among computing nodes. With this idea, a reliability - aware loading balance （RLB） algorithm for data center is proposed. Finally, the experimental results show that the proposed RLB algorithm outperforms than the classic scheduling algorithm ETF and the latest proposed RSEED algorithm. The RLB algorithm can greatly improve the utilization of solar energy with the satisfaction of reliability, job deadline and overall energy consumption budget.

关 键 词：绿色数据中心 可靠性 离散方差 负载均衡 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]