Aydemir, FikriAcar, Züleyha YılmazBaşçiftci, Fatih2020-03-262020-03-2620189.78154E+12https://dx.doi.org/10.1109/UBMK.2018.8566430https://hdl.handle.net/20.500.12395/371833rd International Conference on Computer Science and Engineering, UBMK 2018 -- 20 September 2018 through 23 September 2018 -- 143560Modern Data Centers (DCs) require large amount of power in order to operate their underlying IT infrastructure and the surrounding support systems, such as cooling systems and power conversion systems, due to the complex interdependencies among these support systems. This situation requires a strategy for using the power efficiently while operating a DC. In this study, we firstly propose a software system that serves as data storage for both real-time and offline sensor data about power usage in a DC. Secondly, we provide an analytical platform that uses state-of-the-art deep learning techniques to make predictive data analytics and custom recommendations to DC operators in order to improve power usage effectiveness (PUE) in a DC. The aimed improvement involves estimating the conditions for achieving a PUE rate of 1.03 within an error range of 0.4 percent. © 2018 IEEE.en10.1109/UBMK.2018.8566430info:eu-repo/semantics/closedAccessAnalytical PlatformData CenterDeep LearningPower Usage EffectivenessPredictive Data AnalyticsConference Object6166N/A