Dependability and Sustainability Evaluation of Data Center Electrical Architectures

Camboim, K.; Ferreira, J.; Melo, C.; Araujo, J.; Alencar, F.; MacIel, P.

Faced with the demand to maintain the high availability of data centers (DC), companies are being pressured to seek sustainable alternatives, given that these infrastructures consume a total of 1% of all electricity worldwide [1]. In a time of pandemic (COVID-19), when the digital economy has assumed an even greater share of representativeness, DCs and telecommunications companies need to meet the requisitions of 'everything-as-a-service'. Linked to this are the large amounts of carbon dioxide (CO2) emitted into the atmosphere due to the production and consumption of energy caused by these infrastructures. Given the above, this paper proposes models of energy flow and reliability block diagrams to quantity the environmental impact from different raw materials used to feed the data center loads and computes sustainability and dependability metrics for the entire DC's power infrastructure. According to the specifications for classifying the tiers, this study's hybrid modeling is performed to represent four different electrical architectures. From the model evaluations, we compare whether the availability achieved corresponds to the minimum availability suggested for each tier and show the emissions of CO2 in the atmosphere for each tier over a year. Besides, we apply a parametric sensitivity analysis technique to identify the most critical components for the modeled systems' availability. © 2021 IEEE.