Combined IT and power supply infrastructure sizing for standalone green data centers

• Published in: Sustainable computing informatics and systems Vol. 30; p. 100505
• Main Authors: Haddad, Marwa, Da Costa, Georges, Nicod, Jean-Marc, Péra, Marie-Cécile, Pierson, Jean-Marc, Rehn-Sonigo, Veronika, Stolf, Patricia, Varnier, Christophe
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.06.2021; Elsevier
• Subjects: Computer Science; Distributed, Parallel, and Cluster Computing; Green data center; Infrastructure sizing; Operating Systems; Renewable energy; Renewable energy; Green data center; Infrastructure sizing; Sustainable Computing: Informatics and Systems Renewable energy
• ISSN: 2210-5379
• DOI: 10.1016/j.suscom.2020.100505

•Both IT and power supply sizing for an on-site standalone data center.•Optimal IT sizing algorithm and proved as such, depending on the given workload (tasks and services).•Sizing algorithm to compute a set of green power supply infrastructures consisting of wind turbines, solar panels, batteries and hydrogen system.•Comprehensive experiments to discuss the obtained solutions and justify the influences of different parameters on the data center sizing.•Discussion on the influences of different parameters to help decision maker for choosing the appropriate design of his data center depending on his context, workload and location. In this work, we propose a two-step methodology for designing and sizing a data center solely powered by local renewable energy. The first step consists in determining the necessary IT equipment for processing a given IT workload composed of batch and service tasks. We propose an adapted binary search algorithm and prove its optimality to find the minimum number of servers to handle the IT workload. When the IT sizing is computed, the second step consists in defining the supplying electrical infrastructure using wind turbines and photovoltaic panels as primary sources. Batteries and a hydrogen system are added as secondary sources for short- and long-term energy storage, respectively. In this electrical sizing step, first a set of primary source configurations is determined using a binary search algorithm, then the secondary sources are calibrated so that levels of charge are constant during one day and one year, respectively. Experiments using real IT workload traces and actual meteorological data are conducted to illustrate the provided methodology to decision makers for choosing the best configuration for their data center.