Lawrence Berkeley National Laboratory (LBNL) pioneered using data analytics tools for operational reliability and efficiency at the National Energy Research Scientific Computing Center (NERSC). The data center’s efficient compressor-free cooling system design requires sophisticated control strategies to ensure high availability under all, including extreme conditions. The team built a powerful data collection and analytics platform pulling facility data and supercomputer data together – a rare combination which enables intelligent decision-making in many scenarios. Since 2017, the team started an ongoing commissioning process and installed power usage effectiveness (PUE) metering, saving more than 37% of the non-IT energy consumption – equivalent to 1.8 gigawatt-hours or $100,000 per year.