A Computational Fluid Dynamic Investigation of a Data Centre Employing Rear Door Heat Exchangers

Abstract

As the global demand for data services expands, cooling in data centres continues to evolve towards more efficient and cost-effective systems. Incorporating active rear door heat exchangers has become a popular and reliable method that increases the capability of data centres to operate at higher power densities. This study conducts a thermal analysis of a data centre employing active rear door heat exchangers with the use of computational fluid dynamic (CFD) techniques. The data centre under investigation contains seventy-seven cooled racks with three additional uncooled racks operating in the centre of the hall. The main purpose of this study is to understand how the uncooled racks affect the temperature distribution in the data centre. This study presents a modelling technique which uses temperature and velocity field measurements to facilitate the modelling of rear door heat exchangers. Computer server modelling server was carried out at varying inlet temperature and load. Server simulation results have been utilized with field measurements to create four data centre scenarios. Scenarios were created to show how inlet temperature and load affect the temperature distribution in the data centre. Data centre scenarios have been used to validate and compare with field measurements performed. It was found that heat dissipation in the server was directly related to the server’s velocity profile. From the data centre scenarios created it was found that when higher loaded racks are isolated amongst lower loaded racks the distribution of heat is less significant than if the higher loaded racks were situated in clusters of three or more. It was also found that higher loaded racks could be positioned strategically to diminish the effect of the untreated air produced by the uncooled racks in the data centre. The findings from this paper help to understand the thermal behaviour in data centres and suggests areas to consider when reviewing pre-existing data centre designs.