GreenCompute – Center for Saving Energy in Large-Scale Data Analysis and HPC
The Interdisciplinary Center GreenCompute sets out to research, establish, manage and communicate energy-reducing measures in large-scale computing and in HPC at Humboldt-Universität zu Berlin. It will be based on two pillars - research track and transfer track - which will interact continuously: one dedicated to original, interdisciplinary research, and one concerned with the introduction of technologies and infrastructures which in turn enable and support this research and at the same time implement its findings technically.
The work of the center will be closely coordinated with the novel HPC@HU service and within the Berlin University Alliance (BUA). A cornerstone of the center is the interdisciplinary cooperation of experts in data analysis / HPC with scientists relying on such services. Both parties are represented equally in the pair of speakers and in its PIs. The center will also reach out to a rich network of other organizations that are also concerned with the ever-increasing energy hunger of digital research methods.
Background
Large-scale data analysis and high-performance computing (HPC) are becoming key technologies for all fields of science. However, such processing also consumes a considerable amount of electrical energy, produces an increasingly large CO2 footprint, and becomes increasingly expensive. Research organizations, including Humboldt-Universität zu Berlin, must take significant efforts to mitigate this effect by reducing energy consumption without constraining research progress or quality of services.
This goal must be approached from multiple perspectives, including reduction of computational idle times, raising awareness within the user community, reduction of redundant re-computations, development and rollout of energy-efficient analysis methods and compute units, low-energy configuration of available systems, proper sizing, and forecasting of new systems. While some measures already exist today and are not yet widely used, others require further significant research.
Key Contributors
Speaker: Prof. Dr. Ulf Leser, Department of Computer Science
Speaker: Prof. Dr. Claudia Draxl, Department of Physics
Co-applicant: Malte Dreyer, Director of Computer and Media Service
Coordinator: N.N.
Collaborator: Dr. Maik Bierwirth, Project Development, Computer and Media Service
Further Co-applicants / PIs
Prof. Dr. Sonja Greven, Economics (Statistics)
Prof. Dr. Patrick Hostert, Department of Geography
Prof. Dr. Roland Meyer, Department of Slavic and Hungarian Studies
Prof. Dr. Henning Meyerhenke, Department of Computer Science
Prof. Dr. Susanne Schreiber, Department of Biology
Prof. Dr. Andrea Walther, Department of Mathematics
Prof. Dr. Matthias Weidlich, Department of Computer Science