A review of top-down models of internet network energy intensity

Abstract

Assessments of digital services wishing to take into account an energy footprint of the Internet typically require models of the energy intensity of the Internet. A common approach to the modeling of the energy intensity is to combine estimates of market surveys of installed network devices on a national or global scale and their related power consumption with the total data volume transported at the same scale. These are commonly referred to as top-down models. In order to be applicable, the system boundaries applied in the modeling of energy intensity need to match the system that delivers the particular digital service. The system boundaries applied in past studies vary considerably, resulting in increased uncertainty in the assessment results. In this text, we firstly reconsider the components of the network relevant for the delivery of general public digital services that should be included in an energy footprint. We then reinterpret the existing models in the light of those altered system boundaries. Secondly, we review existing assumptions about annual reduction of energy intensity and consider the effect of uncertainty in these on the overall estimates of energy intensity. We find that the energy intensity of the general Internet network in the year 2014 varies between 0.05 and 0.6 kWh/GB when top-down modeling is applied which is considerably lower than previous estimates had indicated.