Dynamic cooling of an air-cooled data center under transient workloads

Temporal variations in workload traffic can result in thermal throttling or over-cooling depending on the thermal conditions in data centers. While the former is responsible for damaging IT equipment, the latter is responsible for increasing power consumption by the cooling system. In this study, thermal response and cooling efficiency of an operating data center are investigated under generated workload trends by means of transient Computational Fluid Dynamics (CFD) simulations. Strong recirculation and cold-air leakage indicated by the simulation results reduce cooling efficiency of the data center. Variations of the maximum temperature and Rack Cooling Index Low (RCILO) are monitored for a certain time duration under various workload trends. Numerical simulation results reveal that the RCILO is a key parameter for detecting overcooling in an air-cooled data center and the cooling system should be sensitive to the transient variations of workloads. Recirculation and leakage effects are investigated using newly developed metrics. Numerical simulations revealed that the power consumption by the cooling can be reduced by 10% when dynamic cooling approach was adopted.