DescriptionThis paper presents a detailed analysis of a first-of-kind global survey of high-performance computing centers that actively employ techniques within job scheduling and resource management middleware layers for managing energy and power on production supercomputers.
Our group conducted a series of comprehensive interviews of leading-edge supercomputing centers during 2016 and 2017. The group presented the motivation of the survey center selection, questionnaire details, and a preliminary analysis of the survey results in a previous publication. This paper presents a more detailed analysis of the survey results. The goal is to find commonalities, approaches that are to be developed, and hints and guidelines for other centers to move toward a more energy efficient and power aware management of their compute resources.