Field experiment and numerical simulation for airflow evaluation in a data center with row-based cooling

Abstract

Row-based cooling has been recommended for cooling scenarios with high power densities. Existing studies on row-based cooling have mostly focused on good results, and the effectiveness and limitations of it need to be evaluated to optimize it. In this study, some airflow problems were observed by a field experiment for a data center with row-based cooling. Subsequently, a simulation corresponding to this data center was performed to fully learn and verify some airflow conditions below those expected, such as hot spots in some racks and vortex airflow, where the Supply Heat Index of modules 1 and 2 are 0.37 and 0.31. To further determine the airflow characteristics of row-based cooling, a comparative simulation of airflow distribution of row-based cooling and raised-floor cooling was performed. Finally, the effectiveness and limitations of row-based cooling were presented. The results showed that row-based cooling only decreases the Return Temperature Index by 0.4%. The limitations of it include vortex airflow, hot spots in head racks and adjacent racks of in-row coolers, and interaction between different modules; they are easy to occur in row-based cooling and some improvement suggestions are presented in future work.