摘要
针对服务器冷却系统优化工作主要集中在单一socket且存在预测适应时间长精度低等不足提出基于热平衡机理的多层调度策略,使用新的温度预测机制,快速而准确的预测CPU热量后进行内核级调度与SLS策略使单一socket和各个sokcet中热量均匀分布,从而使冷却风扇的转速分布平均而平稳,以最小的风扇转速提供服务器冷却所需的对流换热而达到冷却系统节能优化,实验表明新方案能在不影响系统性能的条件下显著降低风扇能耗达80%,具有较好的节能效果。
Load dispatching strategy is proposed based on the heat balance mechanism.We use new temperature prediction mechanism,with less expense for accurate prediction of heat,based on the kernel level scheduling so that a single socket uniform heat distribution.In each socket,use the SLS strategy to balance its power consumption.Under the condition of power equilibrium,fine-tune overcome local hot spots appear,finally,the cooling fans' speed average and stable distribution,so as to minimize the fan speed to provide server cooling needed for convection heat transfer.Experiment shows that the new scheme can reduce cooling system power consumption at eighty percent,it is a excellent method.
出处
《湖南人文科技学院学报》
2012年第2期92-97,共6页
Journal of Hunan University of Humanities,Science and Technology
关键词
功耗管理
动态负载平衡
热量预测
多层负载调度
power management
dynamic load balancing
thermal prediction
multi-layer load dispatching