Data center thermal management challenges have been steadily increasing over the past few years due to rack level power density increases resulting from system level compaction. These challenges have been compounded by antiquated environmental control strategies designed for low power density installations and for the worst-case heat dissipation rates in the computer systems. Current data center environmental control strategies are not energy efficient when applied to the highly dynamic, high power density data centers of the future. Current techniques control the computer room air conditioning units (CRACs) based on the return air temperature of the air—typically set near . Blowers within the CRACs are normally operated at maximum flow rate throughout the operation of the data center unless they are equipped with nonstandard variable frequency drives. At this setting the blowers typically provide significantly more airflow than is required by the equipment racks to prevent recirculation and the subsequent formation of hot spots. This strategy tends to be overly conservative and inefficient. As an example air entering a given system housed in a rack undergoes a temperature rise of due to the heat added by the system. The return air control strategy strives to keep the entire room at a fixed temperature. Therefore in a typical data center the CRAC supply temperature, and hence the air entering the racks, is and the CRAC return is . At these settings the CRACs can consume almost as much energy as the computer equipment they are cooling [Friedrich, R., Patel, C.D.,2002, “Towards Planetary Scale Computing – Technical Challenges for Next Generation Internet Computing
,” THERMES 2002, Santa Fe, NM; The Uptime Institute, “Heat Density Trends in Data Processing, Computer Systems and Telecommunications Equipment,” White Paper issued by The Uptime Institute, 2000.]. Experiments conducted by the authors using these CRAC settings show that nearly is consumed by the environmental control system for every of heat dissipated by the computer equipment in the authors’ experimental facility indicating that the energy efficiency of standard data center environmental control systems is poor. This study examines several opportunities for improving thermal management and energy performance of data centers with automatic control. Experimental results are presented that demonstrate how simple, modular control strategies can be implemented. Furthermore, experimental data is presented that show it is possible to improve the energy performance of a data center by up to 70% over current standards while maintaining proper thermal management conditions.