The next frontier for communications networks: power management

  • Authors:

  • Kenneth J. Christensen;Chamara Gunaratne;Bruce Nordman;Alan D. George

  • Affiliations:

  • Department of Computer Science and Engineering, University of South Florida, 4202 East Fowler Avenue, ENB 118, Tampa, FL 33620, USA;Department of Computer Science and Engineering, University of South Florida, 4202 East Fowler Avenue, ENB 118, Tampa, FL 33620, USA;Energy Analysis Department, Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, 1 Cyclotron Road, 90-4000, Berkeley, CA 94720, USA;Department of Electrical and Computer Engineering, University of Florida, P.O. Box 116200, 327 Larsen Hall, Gainesville, FL 32611-6200, USA

  • Venue:
  • Computer Communications
  • Year:
  • 2004

Quantified Score

Hi-index 0.24

Visualization

Abstract

Storage, memory, processor, and communications bandwidth are all relatively plentiful and inexpensive. However, a growing expense in the operation of computer networks is electricity usage. Estimates place devices connected to the Internet as consuming about 2%, and growing, of the total electricity produced in the USA-much of this power consumption is unnecessary. Power management is needed to reduce this large and growing energy consumption of the Internet. We see power management as the 'next frontier' in research in computer networks. In this paper, we propose methods for reducing energy consumption of networked desktop computers. Using traffic characterization of university dormitory computers, we show that there is significant idle time that can be exploited for power management. However, current Ethernet adapters in desktop computers lack the capabilities needed to allow existing system power management features to be enabled. We address this problem with a proxying Ethernet adapter that handles routine network tasks for a desktop computer when it is in a low-power sleep mode. This proxying adapter can allow existing power management features in desktop computers to remain enabled and have the computer be 'on the network' at all times. The energy that we expect can be saved is in the range of 0.8-2.7 billion US dollars/year.