Data center characteristic temperature signatures and SHI correlation to nondimensional parameters

Design of high power data center with cost effective reliable cooling solutions is essential for reliability. Over designed cooling systems and rack layouts allowing mixing of air streams increase the energy consumption and cost of data centers. Air current mixing is a factor to control when looking to improve the thermal cooling performance in large data centers. In this work, real time temperature data is collected from sensors located at the input and output of racks. Data collected reveal consistent temperature signatures. Temperature signature trends provide an insight of the mixing of hot and cool air streams in the data center. Uniform temperature trends indicate balanced regions within data center that are not perturbed by variations in air currents. Linear temperatures with monotonic variation are indicative of gradual mixing of hot and cold air in the input region of racks. Presence of pockets of hot or cold air is manifested as "u" shape trends. They are symptomatic of larger flow reversals occurring in the input region in front of racks, with great potential of reducing the coefficient of performance of data centers. The slope temperature in any temperature trend is another indication of the air mixing degree. The supply heat index (SHI) related to the degree of mixing is correlated to parameters of the data center. A nondimensional parameter G is defined in function of Eckert and Reynolds number. G shows a strong linear correlation with SHI for row of racks and a more scattered relation for single racks. G may be used as a first approximation to predict the cooling performance of a data center.