基于芯片温度感知负载调度的数据中心能耗优化

Chip Temperature-aware Workload Scheduling-based Power Consumption Optimization of Data Center

数据中心的高能耗日益成为限制其快速发展的瓶颈问题,通过服务器集群的负载调度可以有效改善数据中心热环境并实现节能降耗。传统的入口温度感知负载调度只考虑了芯片温度过高对服务器工作可靠性的影响,而没有意识到芯片温度过低导致的冷却能耗浪费。基于热循环系数矩阵和服务器传热模型提出的芯片温度感知负载调度策略通过合理优化负载分配避免了服务器的过冷和过热,并降低了数据中心的冷却系统功耗。将芯片温度感知负载调度策略应用到典型的数据中心模型,仿真结果表明:芯片温度感知负载调度策略可以有效实现数据中心节能。

The energy consumption of data center has increasingly become a bottleneck in its rapid development.To improve the energy efficiency of data center,the inlet temperature-aware workload scheduling is commonly adopted to minimize the cooling power,but the over-cooling of server is overlooked.On the basis of the heat recirculation coefficients matrix and the server heat transfer model,this paper proposed a chip temperature-aware workload scheduling strategy which can prevent servers from over-heating and over-cooling by assigning workload intelligently.Meanwhile,the cooling power consumption was also reduced.The chip temperature-aware workload scheduling strategy then was applied to a typical data center model.The simulation result showed that the chip temperature-aware workload scheduling can significantly reduce the power consumption of data center.