FNM-based rack model and CFD-FNM coupling framework for data center rack-level simulation

Data centers, serving as the backbone for today's digital and information-centric world, face the dual challenge of providing robust foundations while contending with significant energy consumption. This paper proposed a novel flow network method (FNM)-based rack model, employing a flow network to capture intra-rack flow pattern. The model extends to construct a corresponding thermodynamic network, enabling the prediction of internal flows of rack and server heat dissipation. Additionally, a CFD-FNM coupling framework, employing a two-way synchronization coupling, is proposed. Through iterative solving, this framework provides a comprehensive depiction of overall flow patterns and temperature distributions within the data center. And the validation with the experiments conducted in the Laboratory of Energy-saving and Low-carbon Technology of Data Center at Xi'an Jiaotong University illustrates the competitiveness of the FNM-based rack model and the CFD-FNM coupling framework in terms of both computation time and prediction accuracy. Finally, through the study of different server/blind arrangement, it is found that: 1) Alternating high and low load servers, with the highest load server placed at the bottom of the cabinet, yields the most effective cooling performance. 2) Placing non-working servers at the top of the cabinet, or replacing them with blinds, improves overall rack cooling efficiency. 3) Increasing the leakage effect of the blind has minimal impact on the rack's outlet temperature, but leads to higher temperatures at the inner server's outlet, thereby reducing cooling efficiency.