Current Internet service-provider networks are typically over-provisioned, with the actual traffic through a network element often being much less than the capacity of the network element. However, current network element power consumption is largely independent of actual traffic. This presents an opportunity to reduce network power usage. Such an opportunity may be exploited locally, by redesigning individual network elements to make them rate-adaptive, or globally, by power-aware traffic routing. Instantiating either approach requires significant engineering effort. We attempt to quantify, as realistically as possible, the power-savings opportunity that can be obtained using these two approaches, in isolation or together. In particular, we investigate whether power-aware routing provides any additional benefit if network elements are rate-adaptive. We adopt a fairly simple model of network power use, where power consumption is attributed to links. A link may be turned off, in which case its power usage is zero. Otherwise, a link that is turned on consumes no less than a certain amount of power, called base power, and power use increases further with traffic. A significant parameter of the model is the ratio of base power over full-capacity power. Since it is difficult to estimate feasible values for this ratio, we investigate multiple scenarios in which its value ranges from 0 to 100%. We demonstrate that the combination of rate-adaptivity and power-aware routing saves a significant fraction of network power consumption, for a wide variety of network topologies, traffic loads, and base power ratios. More specifically, if the base power ratio is 50% or more, then power-aware routing appears to be of significant additional benefit over rate adaptivity alone; if the ratio is 25% or less, then power-aware routing offers a relatively small additional benefit.
[1]
Sujata Banerjee,et al.
A Power Benchmarking Framework for Network Devices
,
2009,
Networking.
[2]
Sergiu Nedevschi,et al.
Reducing Network Energy Consumption via Sleeping and Rate-Adaptation
,
2008,
NSDI.
[3]
TammOliver,et al.
Eco-sustainable system and network architectures for future transport networks
,
2010
.
[4]
N. F. Stewart,et al.
The Gravity Model in Transportation Analysis - Theory and Extensions
,
1990
.
[5]
Moses Charikar,et al.
On non-uniform multicommodity buy-at-bulk network design
,
2005,
STOC '05.
[6]
Andrea Francini,et al.
Performance bounds of rate-adaptation schemes for energy-efficient routers
,
2010,
2010 International Conference on High Performance Switching and Routing.
[7]
Matthew Roughan,et al.
Experience in measuring internet backbone traffic variability: Models metrics, measurements and meaning
,
2003
.
[8]
C. Mas Machuca,et al.
Energy Profile Aware Routing
,
2009,
2009 IEEE International Conference on Communications Workshops.
[9]
Oliver Tamm,et al.
Eco-sustainable system and network architectures for future transport networks
,
2010
.
[10]
Stephen J. Wright,et al.
Power Awareness in Network Design and Routing
,
2008,
IEEE INFOCOM 2008 - The 27th Conference on Computer Communications.