数据中心负载分配对冷却功耗影响的分析和优化  

Analysis and optimization of workload scheduling in data center on the cooling power consumption

作　　者：谷丽君 白焰[1] Gu Lijun;Bai Yan(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China)

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206

出　　处：《低温与超导》2018年第12期67-72,共6页Cryogenics and Superconductivity

基　　金：2016年度山西省科技重大专项(MD2016-02);中央高校基本科研业务费专项资金(2016XS41)

摘　　要：为研究数据中心负载分配对冷却系统功耗的影响,采用基于POD的传热模型描述服务器集群的热环境,并对服务器功耗和冷却系统功耗分别建模,在此基础上提出温度感知负载分配策略以优化冷却系统的送风量和送风温度。将温度感知负载分配策略应用到CFD仿真软件建立的典型数据中心模型,结果表明:在不影响服务器安全运行的前提下,相比于平均负载分配的情况,温度感知负载分配可使数据中心的冷却系统功耗降低4. 9%～9. 0%。In order to study the effect of workload scheduling on cooling power consumption for data center,POD -based heat transfer model was proposed to describe the thermal environment amongst servers.In addition,the server power consumption and cooling power were modeled respectively.Then the temperature -aware workload scheduling strategy was proposed to opti- mize the supply temperature and airflow rate for cooling system.The temperature -aware workload scheduling strategy was ap- plied to a typical data center model simulated with CFD software.The results show that,compared with the case of uniform work- load scheduling,the temperature -aware workload scheduling reduces the cooling system power of data center by 4.9%-9.0% without degrading the server operation safety.

关 键 词：数据中心 负载调度 冷却功耗 CFD仿真 

分 类 号：TP308[自动化与计算机技术—计算机系统结构]