An efficient assessment method for the thermal environment of a row-based cooling data center

Abstract

A suitable thermal environment is essential for the operation of IT equipment in the data center. Thermal environment monitoring methods in data centers are divided into large-scale manual and long-term stationary monitoring. However, long-term stationary monitoring will not effectively capture the changes in the thermal environment within a data center, and large-scale monitoring will result in enormous sensor arrangement costs. Especially for row-based cooling systems, the short air supply paths result in uneven airflow and temperature distribution in the channel, and sensors are needed to capture this information accurately. This study is based on field experiments and numerical simulations, by changing the rack power density, air supply temperature, and air supply flow rate to analyze flow field characteristics. Locations prone to generated hot and cold spots and where airflow mutations are proposed. Then, the correlation between different points and the supply heat index (SHI) was analyzed, and the key monitor point locations were screened. The results show that hot spots in row-based cooling systems are often located in the upper part of racks where in the row terminal and the 2 ∼ 3 racks adjacent to coolers, with the 1.8 m height being the most severe. Cold spots often occur in the height range of 0.7 ∼ 1.5 m in the middle and terminal of the row. The numbers of the new evaluation model’s sensors for the different modules in the data center have only 9 and 4; the R² is 0.824 and 0.819, respectively, and the root mean square error (RMSE) is only 0.012 and 0.019. This method is highly accurate and can be used as a simplified method for large-scale sensor placement and as an alternative to fixed monitoring in data centers.