Toward zero-emission data centers through direct reuse of thermal energy

  • Authors:
  • T. Brunschwiler;B. Smith;E. Ruetsche;B. Michel

  • Affiliations:
  • IBM Research Division, Zurich Research Laboratory, Rüschlikon, Switzerland;IBM Research Division, Zurich Research Laboratory, Rüschlikon, Switzerland;IBM Research Division, Zurich Research Laboratory, Rüschlikon, Switzerland;IBM Research Division, Zurich Research Laboratory, Rüschlikon, Switzerland

  • Venue:
  • IBM Journal of Research and Development
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We have tested hot water data center cooling by directly reusing the generated thermal energy in neighborhood heating systems. First, we introduce high-performance liquid cooling devices with minimal thermal resistance in order to cool a computer system board. This cooling is performed with water at a temperature as high as 608C, thereby eliminating the chillers and their electrical power consumption, and enabling direct reuse of the heat. We collect 85% of the board heat using microscale liquid coolers for CPUs (central processing units), interfaces, and dc (direct current) converters. With our concept, data centers can be cooled in all climate zones throughout the year without a pre-cooled heat carrier. Second, we analyze how the supply of heat and financial payback from customers reduce the total cost of ownership. With 5,000 district heating systems satisfying 9.7% of the thermal demand of Europe, ample opportunities exist for data centers to become heat providers, thereby reducing the associated carbon dioxide emission. Finally, we show how our concept can be developed within 5 years into a zero-emission data center and that such investments are economically viable and ecologically beneficial given increasing energy prices. With such data centers, the IT (information technology) industry can assume a key role in greatly reducing carbon dioxide emissions and global warming by replacing energy-intensive processes with more efficient, digitally assisted processes.