Comparison between conventional and local computer room air-conditioning systems in data center by CFD.

The reduction of energy consumption in data centers is crucial because, recently, the energy consumption of heating, ventilating, and air-conditioning (HVAC) systems has increased owing to huge amounts of data being processed by high-density information technology (IT) equipment. In general, the amount of heat generated by servers and cooling capacity of computer room air conditioning (CRAC) is equal. Meanwhile, the air flow rate of server is higher than that of CRAC. Under these conditions, the high-temperature air, such as the outlet air from the servers may be recirculated to the inlet of the server without being fully captured by the inlet of CRAC. Thus, the energy consumption of CRAC would become high. The advent of cloud service and virtualization technology has changed the operating conditions of IT equipment rapidly. Therefore, it is important to precisely control CRAC systems and servers based on factors such as air flow rate of CRAC, processing location, and server conditions. In this study, the capture efficiency of CRAC (n-index) is defined as the ratio of air volume of server to that of CRAC. Models for estimating indoor air flow were developed based on the thermal equilibrium in data center. Two types of CRAC were assumed, namely, conventional CRAC and local CRAC. Case studies on several heat load and air flow rate of server were examined using CFD to estimate the optimal air flow rate of CRAC based on the n-index of CRAC and allowable temperature of server. Furthermore, the value n of CRAC was investigated based on several processing locations and server conditions.