Numerical Convective Heat Transfer Analysis of a Hybrid-Cooled Data Center Blade Server

Abstract

More powerful and dense data centers require more reliable cooling infrastructures to support them. An increase in energy demand reflects on energy cost, increased greenhouse gases emission and overall cost for upcoming data centers. Therefore, conservation and management of data center energy usage is of great importance. In a typical data center, almost all the electrical power supplied to the IT equipment is converted into waste heat, hence powerful cooling systems are required in order to keep the components below the maximum allowable temperature. For this reason, aside from the IT equipment itself, the highest energy consumer in a data center is the cooling system (more than 30% of the total consumption). Hence, it is important to investigate possible solutions to optimize the thermal management, both by reducing the energy consumption of the cooling systems, and by optimizing the waste heat recovery potential. On this basis, the present paper reports an optimization study carried out in order to achieve energy savings and improved heat recovery potential for a hybrid air-liquid cooled server. In particular, this study investigates the effect of two different baffle designs on the air-cooling efficiency in the server, and on the potential of waste heat recovery at the server outlet. A 2D CFD model is employed to perform a numerical convective heat transfer analysis of the server. Entropy generation minimization (EGM)in conjunction with a multi-objective genetic algorithm (MOGA)are employed to evaluate the cooling system ability to minimize the pump power and unify the air outlet heat stream while keeping the server component temperatures below prescribed constraints.