Studies on airflow and thermal management in data centers.

Abstract

Data centers are computing infrastructure facilities utilized to provide a consistently reliable operating environment for servers, storage and networking devices, generically referred to as data processing equipment. With energy costs rising and information technology (IT) equipment stressing the power and cooling infrastructure, there is now a need to rethink data center strategies, with energy efficiency included to a list of critical operating parameters This project was undertaken with the aim of finding the important parameters required to measure the energy efficiency of a data center. The objectives were achieved through the usage of commercial CFD software, FLUENT, to simulate conditions within the data center and the results were applied to a chosen metric to measure the energy efficiency of the data center. The chosen metrics used were the Return Temperature Index (RTI) introduced by Herrlin, and the Power Usage Effectiveness (PUE) introduced by the Green Grid. The RTI for all cases was above 100%, indicating recirculation effects within the data center under the simulated boundary conditions. This recirculation effect will cause an overall elevation in the return air temperature. The effects of the position of the CRAC return, and the width of the cold aisle were used to check against the energy demands by the data center. It is found that having the CRAC return within the room rather than on the ceiling and widening of the cold aisle width improves the energy efficiency of the data center, as the PUE was reduced. The main reason for the observations is that lower CRAC return temperature was achieved when the return is kept within the room, and when the width of the cold aisle was increased. The observations resulted in a lower energy demand by the CRAC units. The findings suggest that it is possible to bring down the cost of the supporting facility power while maintaining the normal running of the data center without additional cost. It is recommended that future studies include more elements of facility power into the simulation as well as the utilizing the results in the metrics to improve the reliability of the findings.