The 11th International Symposium on Wireless Personal Multimedia Communications (WPMC 2008)

ICT ENERGY CONSUMPTION – TRENDS AND CHALLENGES

Gerhard Fettweis Ernesto Zimmermann
Vodafone Chair Mobile Communications Systems, TU Dresden
Dresden, Germany

ABSTRACT
Many achievements of information society are based on the
Information and communications technology (ICT) systems
global success of information technology which has been
are the core of today’s knowledge based society. Innovations
made possible by innovations in microelectronics. In the last
in this area are adapted at tremendous speed and worldwide
years, ICT systems have been responsible for the enormous
use of ICT has soared in recent years. However, this
economic boom not only of former developing countries like
unprecedented growth comes at a price: ICT systems are
Taiwan, South Korea and Singapore, they have also been the
meanwhile responsible for the same amount of CO2 emissions
source of at least a fourth of the BIP growth of developed
as global air travel. If the growth of ICT systems energy
nations like the United States [4] and the European Union [5].
consumption continues at the present pace, it will endanger
Instead of opening up a "digital divide" between the first and
ambitious plans to reduce CO2 emissions and tackle climate
the third world, the use of ICT has so far lead to the opposite.
change. Increasing the energy efficiency of ICT systems is
thus clearly the major R&D challenge in the decades to come.
II. ICT ENERGY CONSUMPTION TRENDS
The price paid for this enormous growth in data rates and
I. ICT MARKET TRENDS market penetration is a rising power requirement of ICT
systems – although at a substantially lower speed than
Rarely have technical innovations changed everyday life as
"Moore's Law".
fast and profoundly as the massive use of the Internet and
introduction of personal mobile communications. In the past 3
two decades both grew from niche market applications to Electricity cons. of
globally available components of daily life: The first GSM Vodafone network
phone call took place 1991 in Finland – only 15 years later
TWh/year

there were over 2 billion GSM users [1]. In November 2007, 2

every second inhabitant of this planet possessed a mobile
telephone [2]. In the same time span, the number of internet
servers rose by roughly a factor of 1000: from 376’000 to 395
million [3]. The driving force behind these two developments 1
was, and continues to be, "Moore's Law" (or rather the ITRS
roadmap), according to which both the processing power of
CPUs and the capacity of mass storage devices doubles
approximately every 18 months. This in turn renders the use 0
of ever more powerful ICT systems attractive for the mass 2003/04 2004/05 2005/06
market. In order to be able to transport this exponentially
rising amount of available data to the user in an acceptable Vodafone D2 Vodafone Group
time, the data transmission rates both in the (wired) internet
and wireless networks (including cellular, WLAN and Figure 1: Increase in power consumption of Vodafone’s
WPAN) have been rising at the same speed – by about a radio access network over the past years [7].
factor 10 every 5 years, as illustrated in Figure 1.
Short links (1m)
Both in server farms as core units of the internet [6], as well
USB 3.0
as in mobile communications systems [7], a rise of the power
10Gb/s
802.15.3c
WLAN (10m)
consumption of 16-20% per year can be observed in the last
1Gb/s
USB 2.0
UWB intention
years, corresponding to a doubling every 4-5 years, as
802.11vht
100Mb/s
802.11ag 802.11n LTE Advanced illustrated by Figures 2 and 3. As result of this development,
10Mb/s HSPA
server farms meanwhile consume approximately 180 billion
802.11b
USB 1.0 LTE kWh of electricity per year – over 1% of the world-wide
1Mb/s HSDPA
WiMAX Cellular (100m) electricity consumption1. This corresponds to the typical
100Kb/s 802.11
3G R99 / EDGE
yearly electricity consumption of 60 million households –
10Kb/s over a third of the number of households in the EU.
GSM GPRS

1995 2000 2005 2010 2015

1
Based on the data from [6] and a 20% increase for 2006 and
Figure 2: Development of data rates in wireless networks 2007. World wide electricity consumption and production
over time. A ten-fold increase can be observed every five data taken from [8].
years, coinciding with the speed of “Moore’s Law”
The 11th International Symposium on Wireless Personal Multimedia Communications (WPMC 2008)

According to a study by ABI research, the base stations and III. ENVIRONMENTAL AND SOCIETY IMPACT
backhaul networks of the cellular communications networks
The immense influence ICT systems on the power supply
operators consume approximately 60 billion kWh per year
situation can by read from the following 3 facts:
[9], corresponding to roughly 0.33% of global electricity
consumption [8]. Similar figures have been reported in a • Currently, server farms and telecommunications
White Paper of Ericsson [18], which estimates that mobile infrastructure are responsible for roughly 3% of the
networks are responsible for 0.12% of primary energy use world wide electricity consumption. If the present
(electricity typically makes up for roughly one third of growth trend of 16% per year continues, as the
primary energy use). The base stations themselves often have increase of internet traffic2 and the number of mobile
a share of over 80% of the electricity consumed by an phone subscribers suggests3, this consumption rises
operator’s mobile network [7]. Server farms and cellular by a factor of 30 in only 23 years: the current level
networks taken together consume as much electricity as of world electricity consumption. The doubling of
produced by 27 power stations with an output of 1 Gigawatt consumption predicted by the WEC for 2050 [15]
each, or nearly 10% of the electricity produced by nuclear would thus be reached already 20 years earlier, in
power stations world wide in 2005 [8]. This is equivalent to 2030, due to server farms and telecommunications
CO2 emissions of approximately 130 millions tons per year, infrastructure alone.
comparable with the CO2 emissions caused by the whole of • 10% of the electricity consumption of North
Belgium [10]. However, in 2005 "only" 27 millions server America, Western Europe and Japan corresponds to
nodes were located in server farms – compared to over 300 roughly 900 billion kWh per year, equivalent to the
million host in the entire internet (today there are nearly 500 amount of electric power produced in the whole of
Mio [3]). The total amount of electricity consumed by the Central and South America4. To spread the
internet (including all routers, switches, etc.) will thus “information technology standard of living” of the
actually lie far above that of the server farms alone. mentioned countries over the whole world would
already today require 40% of the globally installed
electrical power generation capability. In less than
140 World's server electricity 10 years, the entire existing power generation
120 consumption capacity would not be sufficient for this task5. The
power supply problem alone would thus lead to the
100 (incl. cooling) opening of a "digital divide" between developing
TWh/year

countries and the developed world. Already today,

80 more than two thirds of the operational expenses in
cellular networks in India are due to the buying of
60 diesel for running the generators of the cellular base
40 stations. Cellular network operators spend beyond
200 Mio. Euros each year for their electricity bills in
20 Germany alone. Electricity has grown to a cost
factor comparable to the total wages of the engineers
0 who keep the network running.
2000 2005 • According to a study by Gartner research [16], ICT
systems cause 2% of global CO2 emissions already
US Western Europe Asia Pacific RoW
today, corresponding to
o 2/3 of the CO2 emissions of Germany [10]
Figure 3: Increase in power consumption of data centers [6]. o the total CO2 emissions caused by the
international air traffic [16]
If one sums up the energy consumed by the infrastructure of o 1/4 of the CO2 emissions produced by
cellular wireless networks, wired communications (PSTN) passenger cars world-wide [17]
networks, and the internet (outside of and within the server
farms), a portion of 3% of the world-wide electric power
consumptions appears to be a conservative estimate. 2
A 40% annual growth in IP traffic is predicted in [13] and
But this percentage is still relatively low compared with the by Cisco’s Global IP traffic forecast [19]
quantity of electricity consumed by ICT systems in private 3
8 million subscribers are added/month in India alone [14]
households and offices of enterprises, in particular by laptops, 4
North America, Western Europe and Japan produced 8'600
desktop PCs and displays/monitors. In Germany, the power billion kWh of electrical energy in 2005, Central and South
consumption of ICT systems added up to 38 billion kWh America produced 909 billion kWh [8]
already in 2001 – 8% of the total electricity consumption 5
Assuming a 16% growth of the 3% of electrical power
[11]. Due to the strong growth of market penetration and consumption due to server farms and telecommunications
power consumption of ICT systems, this value might infrastructure; the above mentioned countries correspond to
meanwhile lie beyond 10%, as is confirmed by recent studies roughly 13% of the world’ population; a percentage figure
from the UK [12]. which is obviously bound to shrink
The 11th International Symposium on Wireless Personal Multimedia Communications (WPMC 2008)

IV. RESEARCH CHALLENGES AND SOLUTION APPROACHES optimization. Spatially, temporally and spectrally
localized transmission strategies which minimize the
ICT systems are the key to the innovation capability and thus
energy spent to transmit a single information bit will
the economic success of developed and developing nations.
be crucial to achieve high energy efficiency. This
New challenges arise due to the drastic increase in data rates
speaks very much in favour of OFDM-based
and processing power of ICT systems, which require a
PHY/MAC schemes (LTE and beyond) as opposed
massive research effort into the energy efficiency of ICT
to spread spectrum techniques (3G).
systems. To cite Viviane Reding: "To meet Europe's energy
• The development of suitable utility functions, which
efficiency goals by 2020, we need a high growth, low carbon
enable to solve the multi-criteria optimization
economy. Research and rapid take-up of innovative energy
problems arising from the inclusion of energy
efficient ICT solutions will be crucial […] There is a win-win
efficiency beyond spectral efficiency and coverage
situation in which ICT will promote the competitiveness of
as optimization criteria in an efficient and
EU industry while leading the fight against climate change".
transparent way.
A study by the McKinsey Global Institute found that „The
economics of investing in energy productivity […] are very
Only by finding solutions to these challenging questions can a
attractive. With an average internal rate of return of 17%,
continuing growth of ICT be made ecologically – and
such investments would generate energy savings ramping up
economically – sustainable. Tackling these challenges will
to $900 billion annually by 2020“. Research into energy
also allow for a shift in paradigms in how cellular networks
efficient wireless networks is thus not only ecologically, but
operate. Reducing the power consumption of base stations
also economically very relevant.
down to few tens of Watts will enable to build base stations
Relevant research areas for the next years include:
which can be powered solely based on renewable energy
• The use of high temperature electronics and
sources like solar and wind power. Together with wireless
alternative energy backup solutions, instead of
backhaul (e.g. based on microwave links), these sites will be
batteries. Thus, the energy overhead due to air
extremely easy to deploy. This will again help to narrow the
conditioning can be avoided, which takes up a
digital divide between developed and developing nations,
significant part of the overall energy consumption in
where access to energy and telecommunications infrastructure
server farms and cellular base stations [6].
is often a limiting factor.
• Energy aware strategies for distributed computing
and data storage (distributed data centers), to enable Improving the energy efficiency of ICT components up to the
next generation peer-to-peer and content distribution level where energy autarchy can be reached is also crucial for
services at minimum energy cost. a relatively young, but rapidly growing, sector of the ICT
• Adaptive power management based on current traffic industry: sensor networks. Ridding sensor nodes of the need
/ system load in all areas of ICT systems, e.g.: of a wired interconnect and the lifetime and form factor
o In the radio access network, to reduce the limitations which a battery powered device has to accept is a
power invested in “pilot pollution” in empty key enabler for a massive deployment of sensor networks.
cells, under the condition of not Such “smart dust” may then be used in applications in which
compromising on coverage. the stated limitations so far prevented the massive use of
o In data centers using massively parallel sensor nodes: for example in structure surveillance of
computing, which currently often run at full buildings, bridges, industrial facilities and carbon/fiber
performance, even at low system load. reinforced composite materials used in modern planes and
o In electronic devices for cars or in machine and facility surveillance in industry. Thus,
telecommunications, such as complex the WWRF vision of a thousand wireless devices serving
systems-on-chip. each human being by 2017 can be realized.
• New network architectures based on high density
low transmit power microcells and relaying as V. IMPACT ON OTHER SECTORS OF ECONOMY
opposed to low density, high transmit power, macro
cells. The key here is to find a suitable trade-off The solutions developed for increasing the energy efficiency
between the power consumption analog and digital of ICT will provide new tools, technologies and insights
hardware blocks. which can – and should – also be used for further
• Highly efficient power amplifiers for transmit applications. Low power microelectronics can for example
signals with large peak-to-average-power ratio. help building cars that produce less CO2 per km: already
• Programmable, application specific, highly parallel, today, 20% of the value creation in modern automotive
power efficient digital baseband processors, which industry is based on car electronics. It is expected, that this
on the one hand allow for higher flexibility than percentage grows to 40% by 2015. If energy efficiency is not
ASIC hardware accelerators, and on the other hand improved, gasoline consumption due to car electronics will
consume much less power per function than “general rise accordingly – from 0.5 l / km today to more than 1 l / km.
purpose” DSPs and FPGAs. Car manufacturing and industry automation thus constitute
• Physical and MAC layer optimized for maximum the logical next steps to apply energy efficiency paradigms
energy efficiency, also making use of cross-layer learned in ICT to other sectors of economy.
The 11th International Symposium on Wireless Personal Multimedia Communications (WPMC 2008)

REFERENCES
[1] GSM World: [Link]
[2] Reuters: “Global cellphone penetration reaches 50pct”,
Nov 2007, available at [Link]
[3] Internet Systems Consortium, Inc.; Internet server
statistics available at [Link]
[4] Dale W. Jorgenson (Department of Economics, Harvard
University): „Information Technology and the U.S.
Economy“, Presidential Address to the American Economic
Association, January 2001
[5] European Commission: „Effects of ICT on economic
growth“, Brussels, June 2006
[6] Jonathan G. Koomey, “Estimating Total Power
Consumption by Servers in the U.S. and the World”, Feb.
2007
[7] Vodafone Corporate Social Responsibility Reports
2001/02 to 2005/06
[8] Energy Information Administration, available at
[Link]
[9] Source: ABI Research
[10] UNstats, Millenium Development Goal Indicators,
available online at [Link]
[11] Fraunhofer ISI „Der Einfluss moderner
Gerätegenerationen der IuK-Technik auf den
Energieverbrauch in Deutschland bis zum Jahr 2010“, 2003
(in German)
[12] Global Action Plan: “An inefficient truth”, available at
[Link]
[13] Atsushi Ogasawara “Energy Issues Confronting the ICT
Sector”, Science & Technology Trends Quarterly Review,
No. 21, October 2006
[14] Telecom Regulatory Authority of India, Press release
available at
[Link]
[15] Study by the World Energy Council, available online at
[Link]
[16] See [Link]
[17] See [Link]
[18] Ericsson White Paper „Sustainable energy use in mobile
communications“, August 2007
[19] Cisco Visual Networking Index – Forecast and
Methodology, 2007–2012, available online from
[Link]