Ganesh Ananthanarayanan and Randy H. Katz

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2008-114

September 10, 2008

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2008/EECS-2008-114.pdf

Reducing power consumption in the Internet infrastructure is receiving significant attention. We propose three schemes for power reduction in network switches - Time Window Prediction, Power Save Mode and Lightweight Alternative. These schemes are adaptive to changing traffic patterns and automatically tune their parameters to guarantee a bounded and specified increase in latency. We propose a novel architecture for buffering ingress packets using shadow ports. We test our schemes on packet traces obtained from an enterprise network, and evaluate them using realistic power models for the switches. Our simple power reduction schemes produce power savings of 20 to 35% with minimal increase in latency or packet-loss. With appropriate hardware support in the form of Wake-on-Packet, shadow ports and fast transitioning of the ports between its high and low power states, these savings reach 90% of the optimal algorithm’s savings.


BibTeX citation:

@techreport{Ananthanarayanan:EECS-2008-114,
    Author= {Ananthanarayanan, Ganesh and Katz, Randy H.},
    Title= {Greening the Switch},
    Year= {2008},
    Month= {Sep},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2008/EECS-2008-114.html},
    Number= {UCB/EECS-2008-114},
    Abstract= {Reducing power consumption in the Internet infrastructure
is receiving significant attention. We propose
three schemes for power reduction in network switches - Time Window Prediction, Power Save Mode and
Lightweight Alternative. These schemes are adaptive to
changing traffic patterns and automatically tune their parameters
to guarantee a bounded and specified increase
in latency. We propose a novel architecture for buffering
ingress packets using shadow ports.
We test our schemes on packet traces obtained from
an enterprise network, and evaluate them using realistic
power models for the switches. Our simple power reduction
schemes produce power savings of 20 to 35% with
minimal increase in latency or packet-loss. With appropriate
hardware support in the form of Wake-on-Packet,
shadow ports and fast transitioning of the ports between
its high and low power states, these savings reach 90%
of the optimal algorithm’s savings.},
}

EndNote citation:

%0 Report
%A Ananthanarayanan, Ganesh 
%A Katz, Randy H. 
%T Greening the Switch
%I EECS Department, University of California, Berkeley
%D 2008
%8 September 10
%@ UCB/EECS-2008-114
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2008/EECS-2008-114.html
%F Ananthanarayanan:EECS-2008-114