The life cycle assessment of a UK data centre

Abstract

Data centres are high-energy consumers, and historical assessment of their environmental impact has focused largely on energy consumption. Widely adopted assessment methods consider either single issues or do not comprehensively assess links between issues. One exception is the CLEER Model, which compares life cycle energy and greenhouse gas (GHG) emissions of Cloud-based and present-day services. However, there remains the need to verify components for inclusion in a data centre life cycle assessment (LCA), assess quality and quantity of secondary data, benchmark an existing data centre LCA, assess non-Cloud-based services for multiple impacts, and establish facility areas that are sensitive to change. A hybrid approach, combining process-based and economic input output (EIO) data, was used to perform the screening LCA of an existing UK data centre. The study includes the definition of the goal and scope, modelling assumptions, a life cycle inventory, results and interpretation and a sensitivity check. The dominance of the information technology (IT) operational phase to the overall impact and the severity of the impact on human health are concluded. Due to the use of free cooling, the IT-embodied impact is greater than the combined mechanical and electrical operational impact. Electricity production dominates the total life cycle impact; however, the second most significant impact derives from the disposal of metal refining waste products during the manufacture of IT components and electricity distribution networks. The release of carcinogens is one of the largest contributors to the whole life cycle impact and is almost equal in value between the embodied and operational phases. Finally, a sensitivity check found that a Swedish facility optimised for operational energy efficiency with a 1.25-year server refresh resulted in an embodied impact almost double the operational. It was concluded that current LCI data, software packages and project data allow for a sufficiently accurate data centre LCA model. The results support the need to broaden environmental impact reduction to beyond operational energy consumption for cooling and that building environmental assessment methods (BEAMs) should consider more embodied impacts. It is concluded also that three parameters are sensitive to design changes that influence the overall impact: operational energy for the IT equipment, cooling and power delivery; the energy mix; and the amount of IT equipment across the facility’s lifetime. The results present a clear need to monitor life cycle impact, develop further tools to compare different design/operation options and functional units, improve data and develop an LCA-based BEAM.