Carbon footprint of the Internet

Abstract

A network-based model of the carbon footprint of the Internet is presented and used to determine the carbon abatement provided by Internet-based telecommuting and teleconferencing services to replace car and air travel. The model includes DSL, FTTN and PON access technologies, edge and core network architectures and is based upon currently commercially available equipment. We show that carbon emissions of the Internet need to be taken into consideration in order to obtain an accurate estimate of carbon abatement provided by the Internet.

Full text

CARBON FOOTPRINT OF THE INTERNET
Jayant Baliga, The University of Melbourne
Kerry Hinton, The University of Melbourne
Robert Ayre, The University of Melbourne
Rodney S Tucker, The University of Melbourne
A network-based model of the carbon footprint of the Internet is presented and used to determine the
carbon abatement provided by Internet-based telecommuting and teleconferencing services to replace car
and air travel. The model includes DSL, FTTN and PON access technologies, edge and core network archi-
tectures and is based upon currently commercially available equipment. We show that carbon emissions
of the Internet need to be taken into consideration in order to obtain an accurate estimate of carbon
abatement provided by the Internet.
INTRODUCTION
For a number of years, it has been suggested that broadband telecommunications may be able
to reduce the need for business travel, through the use of telecommuting and teleconferencing
(Telecom Australia 1975; Nairn 2007; Mallon et al. 2007; Webb et al. 2008). In the past, the
claimed benefits have focused on savings in travel time, cost and fuel. More recently the relation
between the energy and greenhouse emission cost of travel has renewed impetus to studies of
telecommunications as an alternative to travel. A reduction in business travel would reduce
greenhouse gas emissions (Nairn 2007). However, the capacity of the Internet would need to be
significantly increased to support good quality video conferencing. If Internet capacity is increased,
the energy consumption, and consequently the greenhouse footprint of the Internet will also in-
crease (Baliga et al. 2007). The authors are not aware of any models that provide a quantitative
measure of the increased greenhouse gas emissions from the Internet if it is to support a significant
reduction in business travel.
In this paper, we present a network-based model of the energy consumption of the Internet,
formulated with data from major equipment vendors (Baliga et al. 2007; 2008a; 2008b). We
model the network infrastructure required to service increasing per-subscriber traffic volumes,
including core, metro, and access networks, and take into account energy consumption in
switching and transmission equipment (Baliga et al. 2007). In the access network, we consider
currently used digital subscriber line (DSL) as well as two future high-speed access technologies
– shared passive optical networks (PON) and fibre to the node (FTTN) (Baliga et al. 2008a).
Using this model we estimate the current annual energy consumption of the Internet in Australia
to be about 75 kWh per subscriber, equivalent to 81 kg of CO2-e, at average access rates in the
order of 500 Kbit/s. The model is then used to estimate the increase in energy consumption re-
quired to support increased access rates that would allow standard definition (SD) and high
definition (HD) video-conferencing. The estimations of increased energy, and resultant greenhouse
gas emissions, are compared against savings from telecommuting and teleconferencing.
BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
TELECOMMUNICATIONS JOURNAL OF AUSTRALIA, VOLUME 59, NUMBER 1, 2009 MONASH UNIVERSITY EPRESS 05.1

NETWORK MODEL
A basic IP network, as used by Internet Service Providers (ISPs), can be logically split into three
main layers – the access network, the metropolitan and edge network and the network core, as
shown in Figure 1. In this section, we outline the technologies commonly used in higher capacity
IP-based networks and develop a model for energy consumption of the network. We model each
of the access technologies using specifications of representative commercial equipment, together
with corresponding metropolitan and core network components to estimate the energy consumed
by the entire network.
Figure 1 Schematic of network structure showing the core, metro and access networks.
Our calculations include the energy required to maintain redundant routers and fibre links
for availability, but do not include the energy consumption of data centres or home networks.
We have deliberately modelled a minimalist network, assuming a national-scale operator deliv-
ering a service, and have not included the inter-network gateways used when multiple service
providers share network facilities. In addition, it must be noted that any future network will
need to incorporate a large amount of legacy network equipment, such as IP over ATM over
SDH over WDM. Such overlays in legacy networks often result in large inefficiencies because of
the need for electronic processing to interface between the network layers. This paper therefore
provides a conservative (i.e. under) estimate on the energy consumption of the Internet. To cal-
culate cooling requirements, we assume that for every watt of power consumed in metro and
core networks, another watt of power is required for cooling (Koomey et al. 2004).
ACCESS NETWORK
The access network connects each home to one of the edge nodes in the provider's network.
There are several different technologies in use today, and new technologies are being developed
(Chanclou et al. 2006). We consider three technologies: asynchronous digital subscriber line
(ADSL), passive shared optical networks (PON) and fibre-to-the-node (FTTN). Shown in Figure
2 are the three access technologies that connect homes, which are on the right, to the Ethernet
switch in the edge node, which is shown on the left. Access technologies differ in the means by
which each home is connected to the edge node. In the following we describe the access techno-
logies, and the equipment they use, in greater detail.
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.2

Figure 2 PON, FTTN and ADSL access networks.
The most common technology in use today is ADSL, shown at the bottom of Figure 2. In
ADSL, copper pairs, originally installed to deliver a fixed-line telephone service, are used to also
deliver a broadband service (Chanclou et al. 2006). The telephone service, which uses the band-
width below 3.4 kHz, is left in place and the higher frequency bandwidth is used for high-speed
broadband services. A modem in each home connects to a digital subscriber line access multiplexer
(DSLAM) at the local exchange.
Copper pair-based access technologies, such as ADSL, are limited by usable bandwidth and
reach, so many ISPs have begun installing fibre-based technologies. Fibre to the premises install-
ation most commonly takes the form of a passive shared optical network (PON), which is shown
at the top of Figure 2. In a PON, a single fibre from the network node feeds one or more clusters
of subscribers through a passive splitter (Chanclou et al. 2006). An access concentrator or Optical
Line Terminal (OLT) is located at the local exchange, and serves a number of access modems or
Optical Network Units (ONUs) located at each of the subscribers' premises. Each subscriber
ONU in a cluster connects via a fibre to the splitter, and from there shares the same fibre connec-
tion to the OLT. ONUs communicate with the OLT in a time multiplexed order, with the OLT
assigning time units to each ONU based on its relative demand.
In areas where subscribers are already served by good-quality copper pairs, a hybrid fibre-
to-the-node (FTTN) technology may be used (Chanclou et al. 2006), shown in Figure 2. Dedicated
fibre is provided from a network switch to a street cabinet close to a cluster of subscribers, and
high-speed copper pair cable technologies used for the final feed to the subscriber premises
(Chanclou et al. 2006). A common copper-based access technology used with FTTN is very high
speed digital subscriber line (VDSL). The street cabinet houses a VDSL DSLAM, which connects
to the VDSL modem at each subscriber premises. The street cabinet also houses one or more
ONUs which connect via fibre to the OLT at the local exchange.
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.3

On the network side, each of the access technologies connects to the metro network through
an Ethernet switch. Table 1 lists equipment used in our model for each of the access technologies
as well as the power consumption and capacity of the equipment.
Table 1 Equipment used in access network models (Baliga et al. 2008a).
OVERSUBSCRIPTION
In general, the transmission and routing capacity provided upstream of the access network is
significantly smaller than the product of the number of subscribers and their maximum access
rates. Much Internet and data traffic is of a bursty nature, with subscribers loading a web page
or block of data via the network, and then generating no further traffic for a period of time whilst
reading or processing this data. Network operators take advantage of this high peak-to-average
traffic ratio and share capacity among subscribers through oversubscription. However, as IP
networks are increasingly used for large file transfers and for streaming services such as IP video,
conferencing, and IP telephony, the traffic demand is both much greater and relatively more
constant. This means there is significantly less scope to employ oversubscription.
In our model we use an oversubscription level of 10. For an access rate of 512 Kbit/s per
subscriber, an oversubscription level of 10 corresponds to 51.2 Kbit/s of capacity per subscriber
in the metro, edge and core networks. We define the access rate as the access rate advertised/sold
to the subscriber. Although we set the oversubscription level to 10, we note that this does not
predetermine the results, as the important parameter is the per-subscriber capacity in the network.
The oversubscription level simply translates the capacity an ISP installs to the access rate it sells
to subscribers.
METRO AND EDGE NETWORKS
The metro network aggregates/concentrates the highly fluctuating traffic from the end subscribers
and serves as the interface between the access network and the core network. Typically within
this part of the network, specialised routers control access to different services within the network,
control access rates, provide authentication and security services, and compile statistics for billing.
Edge routers or switches concentrate traffic from a large number of access nodes, and may home
to more than one core router to improve network availability. Figure 3 illustrates key elements
of the metro, edge and core networks used in our model. On the left side of Figure 3 are the
routers and switches used in the metro and edge networks. Table 2 lists equipment used in the
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.4

metro and edge networks as well as the power consumption and capacity of the equipment. In
the following we describe the equipment used in greater detail.
Figure 3 Metro, Edge and Core networks.
In our model, the aggregation is performed by a Cisco Catalyst 6513 switch, which is a large
Ethernet switch. A Cisco Catalyst 6513 switch has 384 Gigabit Ethernet ports, a switching capacity
of 720 Gbit/s and consumes 3.21 kW. The Ethernet switches uplink to Border Network Gateway
(BNG) routers. The BNG routers perform authentication and access control and we use the Cisco
10008 router as our BNG router. The Cisco 10008 gateway router consumes 1.1 kW and has a
full-duplex capacity to support 8 Gbit/s. Finally, a provider edge router connects to the network
core. Provider edge routers groom and encapsulate the IP packets into a SONET/SDH format
for transmission to the network core. As we are modelling a high growth and high capacity
scenario, we use the large Cisco 12816 as a provider edge router. This device is typical of large
routers available from other manufacturers. A Cisco 12816 router consumes 4.21 kW and in
the configuration used in our model has a full-duplex capacity of 160 Gbit/s. In our model the
provider edge routers connect to the core routers by 10 Gbit/s PoS or SDH links. It is possible
that in some situations the provider edge routers interconnect directly. However, in our model
we assume that all routing is handled by the core routers.
CORE NETWORK
The core network usually comprises a small number of large interconnected routers located in
each major city. An example is the Cisco CRS-1 core router. The Cisco CRS-1 is shown on the
right side of Figure 3. These core routers perform all the necessary routing (packet switching)
and also serve as the gateway to neighbouring core nodes. High capacity Wavelength Division
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.5

Multiplexed (WDM) fibre links interconnect these routers and connect to networks of other
operators. Typically these WDM links comprise 10 Gbit/s Packet over SONET (PoS), SDH, or
10 Gbit/s Ethernet. For long-haul links many use 40 Gbit/s PoS or SDH. In our model we assume
core routers are interconnected by 40 Gbit/s PoS or SDH links. A single-rack Cisco CRS-1 core
router consumes approximately 10.9 kW and has a full-duplex switching capacity of 640 Gbit/s.
In today's Internet, packets traverse an average of 12 to 14 routers between source and des-
tination (Mieghem 2006). Most Internet traffic is from the subscriber premises to a web server,
which is often within a few hops of the network core. In our model, subscriber traffic must traverse
three routers to reach the network core so we assume an average of 10 core routers, giving us
an average of 13 routers in total.
WAVELENGTH DIVISION MULTIPLEXED SYSTEMS
The optical data output from the edge and core routers can only be transmitted relatively short
distances and are generally not suitable for wavelength-division multiplexing into a single fibre.
For this reason wavelength division multiplexed (WDM) transport systems are required for the
long-range communications between the edge and core routers as well as between core routers.
In our model we assume the edge routers are within 200 km of the core node. The WDM
terminal systems connecting the edge nodes to the core node consume 811 W for every 176
channels. If the distance between the two terminal systems is greater than 100 km an intermediate
line amplifier (ILA) is required and consumes 622 W for every 176 channels. In this model the
core nodes are assumed to be an average of 1,500 km apart and so 14 ILAs and 2 terminal systems
are required for every 176 channels.
Table 2 Equipment used in models of metro, edge and core networks (Baliga et al. 2008b).
CARBON EMISSIONS FROM ELECTRICITY AND TRAVEL
In this section we summarise the carbon emissions resulting from electricity consumption, car
travel and air travel. In this paper, the term “carbon emissions” refers to carbon equivalent
(CO2-e) emissions. Carbon equivalent emissions include CO2emissions as well as the global
warming effect of other gases released, such as CH4and N2O (Department of Climate Change
2008).
ELECTRICITY
In recent years there has been a major push toward the production of electricity from renewable
energy sources. However, in the medium term future, the majority of electricity in Australia will
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.6

continue to be generated using brown and black coal, which produces a high level of carbon
emissions (Energy Supply Association of Australia 2008; Department of Climate Change 2008).
The rate of carbon emission per kWh varies across the Australian states because of the different
fuels used, and ranges from 1.31 kg CO2-e/kWh in Victoria to 0.98 kg CO2-e/kWh in Western
Australia and Queensland (Department of Climate Change 2008). In this paper, we assume an
emission rate of 1.08 kg CO2-e/kWh, which is very close to the figure for New South Wales.
CAR TRAVEL
The efficiency of the combustion engines in cars has improved significantly over the past decade.
In addition, low emission hybrid, and all-electric cars are becoming popular, with hydrogen-
powered vehicles projected for the future. However, in the medium term future the overwhelming
majority of passenger vehicles used in Australia will be petrol-based cars (Australian Bureau of
Statistics 2007)
The average rate of fuel consumption of a petrol-based passenger vehicle is 0.112 L/km (11.2
litres per 100 km) (Australian Bureau of Statistics 2007). A petrol-based car produces 2.5 kg
CO2-e per litre of fuel consumed (Department of Climate Change 2008). Combining the average
rate of fuel consumption and the carbon emission rate per litre of petrol consumed, the rate of
emission per kilometre is 0.28 kg CO2-e/km.
AIR TRAVEL
Air travel is believed to be a major source of carbon emissions (Penner et al. 1999). In Australia,
the emission rate of domestic air travel is 0.183 kg CO2-e/km (Wilkenfeld et al. 2002). In addition,
the emissions from aircraft occur in the upper atmosphere and so the effect on global warming
is believed to be 2.2 to 3.4 times greater than carbon emissions from other human activity (Penner
et al. 1999). To account for the increased greenhouse effect of carbon emissions at high altitude
we include a factor of 2.8 (half way between 2.2 and 3.4) and assume the emission rate is 0.512
kg CO2-e/km.
CARBON EMISSIONS OF THE INTERNET
In this section we estimate the carbon emissions of the Internet with ADSL in the access network
and then with FTTN or PON in the access network.
ENERGY CONSUMPTION OF THE INTERNET
Figure 4 is a plot of the electricity consumed per year by an Internet subscriber against the access
rate provided to the subscriber. Included in the plot are the energy consumption of the access,
metro, edge and core networks and the total energy consumption of the network. On the right
vertical axis are the CO2emissions (CO2-e) per year. We have assumed an oversubscription rate
of 10, which corresponds to 10% of subscribers simultaneously using the Internet at full capacity.
The current capacity of the Internet corresponds to an access rate of 0.5 Mbit/s. If used for video
conferencing this would only provide low quality video. For video conferencing with standard
definition (SD) quality, the access rate would need to be increased to 2 Mbit/s. At an access rate
of 0.5 Mbit/s the network consumes 75 kWh of electricity per year per subscriber and results in
81 kg of CO2-e. If the access rate were to be increased to 2 Mbit/s, the network would consume
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.7

80 kWh of electricity per year per subscriber and result in 86.4 kg CO2-e. This increase from 75
kWh to 80 kWh is quite modest because the per-subscriber energy consumption of the network
at low access rates is dominated by the energy consumed in the access network, in particular the
subscriber’s access modem. The energy consumption of this modem is almost independent of
access rate. The small increase in energy consumption arises from the increased transmission and
switching capacity required in the metro, edge and core networks. Thus as a consequence of
upgrading network capacities to deliver SD quality video conferencing to each subscriber, the
carbon emissions of the Internet, assuming ADSL in the access network, would increase by 5.4 kg
CO2-e per-subscriber or 6.7%.
Figure 4 Per household/subscriber electricity consumption and CO2-e emissions of the network with
ADSL in the access network.
ENERGY CONSUMPTION OF ACCESS NETWORKS
As shown in Figure 4, energy consumption in the network is dominated by the consumption in
the ADSL access portion of the network. Thus it is appropriate to consider alternative broadband
access technologies on the basis of capacity and energy consumption. The theoretical maximum
attainable access rate with ADSL2+ is 24 Mbit/s. However, the theoretical maximum degrades
with distance and so the full access rate is attainable by only a small percentage of the population.
FTTN and PON access networks are capable of delivering an access rate in excess of 10 Mbit/s
to all subscribers they serve. Figure 5 is a plot of the electricity consumed per year against access
rate with ADSL2+, FTTN/VDSL and PON in the access network. On the right vertical axis are
the CO2emissions (CO2-e) per year. We have again assumed an oversubscription rate of 10. At
an access rate of 2 Mbit/s, a network with a PON consumes 73.5 kWh of electricity and results
in 79.4 kg CO2-e per year. For the same access rate, a network with FTTN consumes 149 kWh
of electricity and results in 161 kg CO2-e per year.
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.8

Figure 5 Per household/subscriber electricity consumption and CO2-e emissions of the network with
ADSL, FTTN or PON in the access network.
For high definition (HD) video conferencing, an access rate of 10 Mbit/s is required. To
provide an access rate of 10 Mbit/s, a network with FTTN would consume 176 kWh of electricity
and result in 190 kg CO2-e per year. In comparison, a network with PON providing an access
rate of 10 Mbit/s only consumes 101 kWh of electricity and results in 109 kg CO2-e per year.
Table 3 summarises the carbon emissions of the Internet when ADSL, FTTN and PON are used
in the access network at current access rates (0.5 Mbit/s) as well as access rates suitable for video
conferencing – 2 Mbit/s (SD) and 10 Mbit/s (HD).
Figure 5 indicates that to provide HD video conferencing using FTTN in the access network
the electricity consumption and carbon emissions would need to increase by 135%. If PON was
used in the access network the electricity consumption and carbon emissions would only need
to increase by 35%.
Table 3
Carbon emissions per subscriber attributed to the Internet with ADSL, FTTN and PON when the access rate is 0.5 Mbit/s (current),
2 Mbit/s (SD) and 10 Mbit/s (HD).
CARBON EMISSION SAVINGS FROM THE INTERNET
In this section we analyse the carbon emissions savings that would arise from use of broadband
leading to reduced car and air travel. We analyse SD (2 Mbit/s) as well as HD (10 Mbit/s) quality
video conferencing. In both cases we assume an oversubscription rate of 10. If we consider SD
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.9

quality video conferencing, an oversubscription rate of 10 means that during the busiest period
of the week, at most 10% of subscribers are using video-conferencing with SD quality. The re-
maining 90% of subscribers are not using the network in any way, including web surfing. Net-
works are dimensioned to support the busiest period of the week/month. Consequently, the ac-
tual number of hours video-conferencing is performed is irrelevant. In addition, with current
technology, although at most times the network is well below its capacity, the network equipment
remains fully operational and consumes the same amount of power it would during the busiest
periods.
It may seem disproportionate, for example, to compare a 2% reduction in car travel and a
network capacity with support for 10% of homes using video-conferencing. However, it should
be noted that 2% is an average value while 10% is a peak value. For example, if 10% of people
were to telecommute on Fridays, this would only result in a 2% reduction in travel. If this 10%
of people used video-conferencing at the same time on any given Friday the network would be
at full capacity and any additional traffic during this period would result in degraded service for
all users. Additional traffic includes viewing web pages, downloading email, music or video.
The emissions from broadband with each access technology for SD and HD quality video
conferencing is given in Table 3. When calculating savings, we sum the required increase in
broadband carbon emissions from the current level of 81 kg CO2-e with the savings from reduced
car and/or air travel. We note that our results showed that FTTN consumes more energy than
ADSL while PON is more efficient than ADSL. In particular, a network with PON with a 2 Mbit/s
access rate would consume less power than is currently consumed by the network with ADSL at
an access rate of 500 Kbit/s. In the following we analyse the savings that arise from reduced car
and air travel and compare them to carbon emissions from broadband.
CARBON EMISSIONS SAVINGS THROUGH THE INTERNET: CAR TRAVEL
In Australia there are 7.9 million homes with 11.3 million cars (Pink 2008). This equates to 1.4
cars per home. In addition, the average passenger vehicle travels 8100 km to and from work per
year (Australian Bureau of Statistics 2007). When comparing the carbon emissions of broadband
and car travel (to and from work) we assume one broadband connection per home and 11,340
km to and from work per home per year. The carbon emission savings calculated as a function
of percentage reduction in car travel to and from work. Figure 6 is a plot of CO2-e savings for
percentage reductions in car travel from 0% to 10% with each of the access networks for SD
and HD video conferencing. We see that with ADSL and PON with SD quality video conferencing
even a reduction of 0.2% in car travel results in CO2-e savings. However, with FTTN and SD
quality video, a 2.5% reduction in car travel is required to start realising CO2-e savings. The
CO2-e savings are reduced when HD quality video is used with both PON and FTTN.
Nairn analysed the environmental benefits of a 5% reduction in daily commuting travel
(Nairn 2007). If 5% of people were to work from home using broadband, this would result in
an average of 567 km less travel per vehicle and a carbon emissions reduction of 159 kg CO2-e.
Table 4 is a summary of the net saving from reduced car travel for each of the access technologies.
In all cases, using the Internet to work from home does indeed result in significant carbon emission
reductions. If the currently used ADSL network is augmented to support SD video conferencing,
the carbon emission reduction would still be 153 kg. With a FTTN access network, carbon
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.10

emission savings are well below 159 kg CO2-e. This indicates that increased carbon emissions
from the Internet must be accounted for when estimating carbon abatement. We note that using
a PON in the access network results in more than 159 kg CO2-e saving and this arises because
a PON consumes less energy than ADSL.
Figure 6
CO2-e emission savings per household/subscriber for a given percentage reduction in car travel with ADSL, PON or
FTTN in the access network, and SD or HD quality video-conferencing.
Table 4 Summary of carbon emission savings for car travel per home.
CARBON EMISSIONS SAVINGS THROUGH THE INTERNET: DOMESTIC AIR TRAVEL
In this section we analyse the carbon emissions savings that would be realised if broadband res-
ulted in reduced domestic air travel alone. We do not include the carbon emission savings from
reduced car travel calculated in the previous section.
In 2006, approximately 38.4 million domestic air trips were taken annually, resulting in a
total of 46.9 billion kilometres travelled domestically via air (Pink 2008). In Australia there are
7.9 million homes (Pink 2008) so the number of air kilometres travelled per household is approx-
imately 5900 km. The carbon emission savings arising from reduced air travel are calculated as
a function of percentage reduction in domestic air travel per household. Figure 7 is a plot of
CO2-e savings for percentage reductions in air travel from 0% to 10% with each of the access
networks for SD and HD video conferencing. ADSL and PON, with SD quality video conferencing,
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.11

require only a small reduction of 0.2% in air travel to begin realising CO2-e savings. If the access
network is a FTTN network a 2.6% reduction in air travel is required before CO2-e savings are
realised. The CO2-e savings are lower when HD quality video is used with both PON and FTTN.
Figure 7 CO2-e emission savings per household/subscriber for a given percentage reduction in domestic
air travel with ADSL, PON or FTTN in the access network and SD or HD quality video-conferencing.
CARBON EMISSIONS SAVINGS: CAR AND AIR TRAVEL COMBINED
A modest reduction in car and/or air travel will result in net CO2-e savings. Given costs of con-
struction for new cable installations, FTTN is likely to be the most economically viable option
for upgrading Internet in existing urban areas, especially when underground cabling is mandated
(Chanclou et al. 2006). On a per-year operational basis, a FTTN network is capable of producing
net carbon abatement if increased levels of telecommuting and teleconferencing lead to a 1.3%
or larger reduction in car and air travel. Our results counter-intuitively suggest that reducing car
and air travel are equally important. On a per kilometre basis carbon emissions from air travel
are double the emissions from car travel, but the distance travelled to and from work by car per
household is double the distance travelled using domestic aviation. Businesses should put equal
effort into encouraging telecommuting and teleconferencing.
CONCLUSION
We have presented a model to quantitatively estimate carbon emissions of the Internet. Our
model includes the metro, edge and core networks and in the access network we have considered
ADSL, PON and FTTN with VDSL. We estimate that the Internet in Australia currently emits
81 kg CO2-e per-year per-subscriber. Our results confirm the widely held belief that significant
carbon emission savings can be achieved if broadband-enabled services result in reduced travel,
in particular through increased telecommuting and teleconferencing. However, we also show
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.12

that the energy consumption of the Internet will need to increase to support the higher access
rates required by video conferencing. Our results indicate that some access technologies provide
greater abatements than others. We believe that in future, more effort needs to be given to im-
proving the energy efficiency of network infrastructure in the Internet. This will enable optimum
carbon abatement as access rates increase.
REFERENCES
Australian Bureau of Statistics. 2007. ‘Survey of Motor Vehicle Use’. [Internet]. Australian Government.
Accessed 30 September 2008. Available from:
http://www.ausstats.abs.gov.au/ausstats/subscriber.nsf/0/331AFCE3EFECE3EECA257378007BC0
F2/$File/92080_12%20months_ended_31_October_2006.pdf.
Baliga, J; Hinton, K; Tucker, R S. 2007 ‘Energy consumption of the Internet’. Proceedings of COIN-ACOFT
2007. June; Melbourne, Victoria: 1–3.
Baliga, J; Ayre, R; Sorin, W V; Hinton, K; Tucker, R S. 2008a. ‘Energy consumption in access networks’.
Proceedings of the Optical Fibre Communications (OFC) Conference. February; San Diego, California:
1–3.
Baliga, J; Ayre, R; Hinton, K; Sorin, W V; Tucker. 2008b. ‘Energy Consumption in Optical IP Networks’.
submitted to the Journal of Lightwave Technology.
Chanclou, P; Gosselin, S; Palacios, J F; Alvarez, V L; Zouganeli, E. 2006. “Overview of optical broadband
access evolution,” Institute of Electrical and Electronics Engineers Communications Magazine
(August): 29–35.
Department of Climate Change. 2008. ‘National Greenhouse Accounts (NGA) Factors’. [Internet]. Australian
Government. Accessed 30 September 2008. Available from:
http://www.climatechange.gov.au/workbook/pubs/workbook-feb2008.pdf.
Energy Supply Association of Australia. 2008. ‘The Energy Industry: Facts in Brief’. [Internet]. Energy Supply
Association of Australia. Accessed 30 September 2008. Available from:
http://www.esaa.com.au/the_energy_industry:_facts_in_brief.html.
Koomey, J; Chong, H; Loh, W; Nordman, B; Blazek, M. 2004. “Network electricity use associated with
wireless personal digital assistants,” American Society of Civil Engineers Journal of Infrastructure
Systems (10): 131–137.
Mallon, K; Johnston, G; Burton, D; Cavanagh, J; Burton, B. 2007. ‘Towards a High-Bandwidth, Low-Carbon
Future’. Climate Risk Pty Limited (Australia). Fairlight, New South Wales.
Mieghem, P V. 2006. Performance Analysis of Communications Networks and Systems, 358. Cambridge:
Cambridge University Press.
Nairn, R J. 2007. ‘Broadband telecommunications and urban travel’. Telecommunications Journal of Australia
57 (2-3): 26.1–26.9. DOI: 10.2104/tja07026.
Penner, J E; Lister, D H; Griggs, D J; Dokken, D J; McFarland, M. 1999. ‘Special Report on Aviation and
the Global Atmosphere’. [Internet]. Intergovernmental Panel on Climate Change (IPCC). Accessed
30 September 2008. Available from: http://www.ipcc.ch/ipccreports/sres/aviation/index.htm.
Pink, B. 2008. ‘2008 Year Book Australia’. Australian Bureau of Statistics No. 90. Canberra: Commonwealth
of Australia.
Telecom Australia. 1975. “Telecom 2000 – An exploration of the Long-Term Development of
Telecommunications in Australia. Australian Telecommunications Commission.
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT 05.13

Webb, M., et al. 2008. ‘SMART 2020: Enabling the low carbon economy in the information age’. The
Climate Group. London.
Wilkenfeld, G., et al. 2002. ‘Australia’s National Greenhouse Gas Inventory 1990, 1995 and 1999’. Report
to the Australian Greenhouse Office. Canberra: Australian Greenhouse Office.
Cite this article as: Baliga, Jayant; Hinton, Kerry; Ayre, Robert; Tucker, Rodney S. 2008. ‘Carbon footprint
of the Internet’. Telecommunications Journal of Australia. 59 (1): pp. 5.1 to 5.14. DOI: 10.2104/tja09005.
CARBON FOOTPRINT OF THE INTERNET BROADBAND FOR THE SUSTAINABLE ENVIRONMENT
05.14