Abstract
Growth of data centers’ demand for electricity with respect to the energy cost emerges the need for more effective cooling solutions. This requirement necessitates a comprehensive assessment of each cooling technology in terms of their energy efficiency and economic performance. Here we developed a comparative analysis for four different cooling methods for a 1.7MW data center to indicate how the type and scale of the cooling technology impacts the total efficiency and costs. IT load is kept constant among all the four studies and only the cooling infrastructure configuration has been changed. Both air and hybrid (air-liquid) cooling are examined and for the liquid-cooled section, liquid-to-liquid (L2L), and liquid-to-air (L2A) coolant distribution units (CDU) are assessed at two levels of rack and row scale. Total Usage Effectiveness (TUE) and Total Cost of Ownership (TCO) results are presented for all different designs. The TCO calculations are divided into some subsections where costs for the primary and secondary side of the data center are specified. The capital cost for equipment, cooling units, preparation inside and outside of the data center, plumbing, installation, commissioning, and servicing etc. are considered while preparing the total cost for a particular cooling technology. Here, we considered N+1 redundancy for the cooling equipment on the secondary side and UPS system. The operation cost calculation includes the energy usage of the components inside and outside of the data center. Comprehensive analysis has been performed while transitioning from air cooling to hybrid cooling as efficiency of the components play an important role in determining the total power consumption. In this analysis, TCO modeling is facilitated with TUE number for cases being studied and power consumption from auxiliary components are included in the analysis. Results demonstrate a drop of 8.5% and 3.4% in total cooling energy usage and operating costs respectively due to the transition from air towards hybrid cooling. In the next step, the sensitivity of the TCO model to the type of coolant fluid and relevant thermohydraulic properties has been evaluated and presented. In the concluding phase, an optimization study was undertaken to illustrate the appropriate utilization of liquid cooling in order to achieve a design that is more economically efficient. This analysis aims to provide insights regarding the technological opportunities for further power and cost reductions for data center owners and highlight the most impactful parameters on both operational and capital costs. The outcome of this research demonstrates the actual cost per MW of IT which is a simple yet effective metric for each cooling architecture.
