On cold-aisle containment of a container datacenter

Abstract

In this study, influences of various blockage arrangements as well as cold-aisle containment on the overall efficiency of a container datacenter are reported. The test container datacenter contains ten racks and the cold air is supplied from drop ceiling. The experiments are performed with each rack equally delivering a power of 3 kW. Eight different layouts are examined. Appreciable hot air recirculation prevails especially at the entrance of cold aisle. Adding blockage plates to the computer room air handler (CRAH) at the entrance of computer racks only improve marginally the associated rack cooling index (RCI). Layout with enclosing the end of cold aisle while let open the entrance of cold aisle also performs poorly. Full containment of the cold aisle with high jet air flowrate from the supplied grilles gives the worst performance. Layout with enclosing at the entrance of cold aisle but let open the end of cold aisle shows a much superior temperature distribution and a rather high RCI of 99%. Higher supplied flow may induce jet airflow pattern, causing the Coanda effect to result in fluctuating behaviors of RCI and SHI alongside the cold aisle (zigzag phenomenon). A lower supplied air flowrate without Coanda effect for a full containment design give the best overall performance. (C) 2016 Elsevier Ltd. All rights reserved.