Effect of rack modeling detail on the numerical results of a data center test cell

Abstract

This paper is a continuation of the study [1] that describes the comparison of numerically-predicted and experimentally-measured temperature distributions in a small data center test cell. The data center cell has a floor area of 900 ft <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and comprises one rack, one air-conditioning unit, and perforated tiles. The Computational Fluid Dynamics (CFD) results in study [1] seem to over-pronounce the hot and cold spots and under-predict the mixing between the hot and cold air streams relative to measurements. One of the possible reasons for the disagreement was believed to be the simplified representation of the server rack in the CFD model. This study investigates the effect of server rack model detail on the prediction of temperature distribution within the data center test cell. It is found that insufficient rack-modeling detail is not a likely cause of the poor agreement between experiment and prediction. However, it is demonstrated that rack modeling detail can significantly impact predictions in certain applications.