| Data centers are mission critical facilities involving high capital expenditures and are designed to operate with little or no downtime. Increase in computing power resulting from high performance microprocessors, packages, and modules and the deployment of high heat-load computer racks in high density configurations, has escalated the thermal challenges in today's data center systems. The amount of energy spent in cooling the server heat loads depends on the data center cooling design. In a well-managed data center, for every watt of a server power, one extra watt is consumed by coolers, UPSs, PDUs etc. The optimization of cooling design can significantly lower the operational cost of a data center facility.;This dissertation addresses the issues of thermal inefficiencies and establishes a set of design guidelines for thermal management of data centers. The effectiveness of seven different data center configurations is studied and compared. The configurations studied include different combinations of raised floor and ceiling supply and return vent locations subject to specific constraints. The parametric study for ceiling height, tile flow rate, and the location of return vents was performed. The use of ANOVA (Analysis of Variance) method is discussed for the significance of different parameters on the thermal performance of these data centers.;Numerical methods are widely used to model existing and new facilities. Validation of existing numerical techniques is an important step in facilitating good thermal design of data centers. An experimental-numerical validation for a large real-world data center facility is presented.;A software tool using the Neural Network (NN) method has been developed for the real-time prediction of rack cooling performance for clusters in a simple room environment. The Neural Network models have been trained on thousands of CFD runs. A good overall accuracy is achieved by using the Neural Network approach. Because of the real-time nature of the calculations, the NN approach readily facilitates optimization studies. Example cases are discussed, which show the integration of the NN approach and a genetic algorithm used for optimization.;The major scientific contributions of this dissertation are the quantitative comparison of the thermal performance of seven different data center airflow duct designs, the effect of ceiling height, tile flow rate, location of return vents on the thermal performance of the seven data center types, experimental-numerical validation of a large real-world data center facility, the methodology based on the regression technique for the prediction of cold-aisle end airflow boundary conditions, the methodology based on the Neural Network technique for the real-time prediction of rack cooling performance, and the methodology using Genetic Algorithms (GA) in combination with the Neural Network cooling-prediction engine for the optimization of cluster layouts. |
