Master in Telematics Engineering

Polytechnic University of Cataluña

XSBA-2012

Energy efficiency in networking

Concepts involved, strategies

1. - Justification.

During the past 30 years, the growth of telecommunications technologies (IT) has been exponential.
Moreover, in recent years, the wide dissemination of broadband communications, mobile phones, etc,
make this growth model not scalable in terms of energy.

Different organizations have published reports warning about this fact. For example, the Carbon Trust
(an organization founded by the UK government), published a study stating that the office equipment
currently accounts for 15% of total energy consumption of a company, and is projected to increase to
30% in 2020 if not action is taken. Globally, it is estimated that the ICT producing between 2 and 4% of
global emissions of greenhouse effect gases. A significant part of these emissions, sixth part, is due to
telecommunications networks.

Just as the European Union predicts that the power consumption in data centers will double between
2007-2020 reaching 104 TWh.This organization leads an initiative called the EU Code of Conduct for
Data Centres pursuing that all agents get involved in measuring consumption and implementing
energy efficient technologies.

The predictions regarding household consumption are not expected promising .It is expected by 2015
that electricity consumption increased 50 TWh annually. Because of this, the EU developed the EU
Code of Conduct Broadband Equipment, containing principles and actions to limit consumption without
affecting the technological advances and the use of communication networks at home.

In a context of scarce energy resources, with a consequent increase in price, this predicted scenario
has motivated that large technology companies (Google, Microsoft, Amazon, Yahoo, etc.) are taking
steps to reduce the power consumption of their data centers and telecommunications infrastructure.As
an example, it deserves be mentioned the Google data center in Hamina (Finland) that takes profit of
the low temperatures and the water of the sea to cool his facilities.

Similarly, the concern about climate change, makes possible the development of new
telecommunication technologies. It is expected that they have the potential to eliminate the estimated
15% of the "ecological footprint".This report will try to describe the concept of energy efficiency applied
to telecommunications technologies (IT), good and bad practices and main strategies taken in different
parts of the network.

References:
*Six drivers for the development of Green IT.-Carbon Trust – Sunday Telegraph (oct 2012).
*The Energy Eficient Network.-Mark Blowers. The Butler Group (January 2009)
*Power Consumption in Telecommunication Networks:Overview and Reduction Strategies – W.Vereecken and others – IEEE
Communications Magazine, vol 49 issue 6.
*The next frontier for communications networks: power management – Kenneth Christensen, Chamara Gunaratne, Bruce Nordman, Alan
D.George - Computer Communications 27 (2004) 1758–1770

2.Analysis Of The Distribution Of Consumption In Telecommunications Networks.

A telecommunications network can be segmented into different components, which affect the
overall power consumption in different ways. For our analysis we divide the network into
three major entities (with their respective components):

Figure 1.Network Telecommunication [1]

- End-user.
- Access Networks
- Backbones.

Within the user equipment it is basically noted the DTE (Data Terminal Equipment),
especially the personal computers of the users. The DCE (data circuit-terminating
equipment) ADSL modems located primarily in homes, it is mentioned in the access
networks.

Although not the main subject of this report, we quote it because it is the component that
produces the highest energy consumption. In 2000, it was estimated that, excluding
residential, 9% of electricity consumption was attributable to IT equipment, about 97 TWh per
year. Of this consumption, approximately 54.8% was attributable to PCs, servers and
monitors. Excluding monitors, we could say that 31 TWh / year in PCs and 12 TWh / year in
servers. Studying the pattern of use of a PC is concluded that the energy consumption in a
PC is highly inefficient and could be implemented multiple measures to reduce it (stand-by,
turn-off etc).

3. Network Efficiency In The End-User.

Despite all the efforts that can be made by researchers, scientists, engineers etc. in order to
make more efficient components, proposing new protocols, network architectures, paradigms
of consumption, an important part of the electricity consumption (possibly the most important,
since their composition may influence the other components) is the end user.

[1] Willem Vereecken, Ward Van Heddeghem, Margot Deruyck, Bart Puype, Bart Lannoo, Wout Joseph, idier Colle, Luc Martens, and
Mario Pickavet “Power Consumption in Telecommunication Networks: Overview and Reduction Strategies” Ghent University - IBBT,
Department of Information Technology (INTEC) in Advanced Networks and Telecommunication Systems, 2008.
Often, the end user makes an increasing-use of resources. In any case it is proposed a
penalty or any kind of coercion in the customs and habits of consumers. But what is certain is
that more responsible behavior could result in a reduction in electricity consumption and
therefore increased networking efficiency.
Even the tendency of users to increase the use of consumer-goods (in this case the energy)
when the price of these decreases (a phenomenon called Khazoom-Brookes Postulate or
Jevons paradox) can make useless all the achievements in research and development.

Thus, it is mandatory in the user equipment, how to mitigate their consumption and use it
more efficiently.

The first aspect is mentioned is the use-paradigm of the personal computer. In 2000, world
energy consumption in PCs was 31 TWh, while in 2005 it had grown to 46 TWh. But how
much of this consumption is productive? In other words, the use that is given to a personal
computer, is it efficient?

figure 2. Energy Consumption in commercial PC [2]

This figure shows a typical pattern of use of a personal computer for business use. It can be
seen as active use is minimal (about 5% of the time), while most of the time, the computer
consumes energy without doing anything productive. Thus saving opportunities are evident.
Most of the time the computer is doing nothing productive (idle), could be in sleep mode.

figure 3. Potential Reduction of consumption [2]

[2] Reducing the Energy Consumption of Networked Devices. – Bruce Nordman, Ken Cristensen – IEEE 802.3 tutorial, july 19,2005

In this sense, the Environmental Protection Agency and the Department of Energy of the
United States has developed an initiative called Energy Star.It promotes various initiatives in
the direction of lowering the power consumption of PCs:
 -Power Management and PC monitor.
-Efficiency power supply.
-Limits consumption in idle mode.
-Sleep-mode connectivity.

It can be seen, in a clearly way, that a behavior change in users, could grow their equipment
efficiency and significantly reduce the consumption of energy. As for the network equipment,
there are initiatives and technologies that increase efficiency.

For example, power management ADSL2 implements different three modes of operation.
One can observe the energy saving compared to a standard ADSL line.

figure 4. Adsl Vs Adsl2 [2]

Other initiatives from the side users can be taken in order to reduce such electricity bill, as
might be the most efficient design of protocols, using components that "turn off" parts that
are to be used, power management systems for computers "centralized "in offices, use of
sleep modes and even automatic shutdown of equipment.

In the user switches can be used a kind of dynamic power management, that is, turning off
the core (maintaining buffers powered) between frame reception, forecasting long times
without receiving frames and power-up the core when receiving frames.

Basically, the measures result from the idea of not feeding unneeded parts that are not being
used, and may be different component units, reception, protocols, etc. These initiatives,
combined with the behavior of users and system administrators can increase the efficiency of
this part of the network, reducing the electricity bill and increasing the environment
protection.

[2] Reducing the Energy Consumption of Networked Devices. – Bruce Nordman, Ken Cristensen – IEEE 802.3 tutorial, july 19,2005
[3] Energy-Efficient Next-Generation Networks (E2NGN)- Pulak Chowdhury-B.S. (Bangladesh University of Engineering & Technology)
2002 M.S. (McMaster University, Canada) 2005
[4] Reducing the Energy Consumption of Networked Devices. – Bruce Nordman, Ken Cristensen – IEEE 802.3 tutorial, july 19,2005
[5] Energy Star – www.energystar.gov
4. Power Consumption In Access Network

The access network is the network whose mission is to provide end users with a connection
through which Internet access. Its structure is a tree, connecting users to a central node in
which the different links are added and connected to the backbone. Due to the heterogeneity
of existing technologies, each one, has a different level of aggregation, which is key to
measuring the consumption per user. To more clearly between different kinds of access
networks take as a basis the following figure:

Figure 5. Access Networks [6]

Quickly we can distinguish between two kinds of access networks, wired access networks
(HFC, DSL, FTTN, PON) and wireless access networks (WiMAX and UMTS) these are not
the only technology of existing access networks but are the most widely used by service
providers. The energy consumption in each of these access networks can be divided into
three components:

 Customer premises equipment (i.e., the modem)

 Remote node or base station (base transceiver station, BTS),
 Terminal Unit (Located In the Central office)

We will use in most cases for our examples the more technologies implemented by service
providers, since in recent years around the world has been taken as a metric design the
energy consumption, since previously only to under the cost metric for the service provider.
[6] Jayant Baliga, University of Melbourne and National ICT Australia Robert Ayre, Kerry Hinton, and Rodney S. Tucker, University of
Melbourne “Energy Consumption in Wired and Wireless Access Networks” IEEE Communications Magazine • June 2011 71

The energy consumption in all access networks mentioned above can be expressed by the
following form:
 PRN 1,5 PTU
( Power Comsumption Per Customer=Pa=PCPE+ NRN +
NTU )
[2]
Where ,
PCE=Power consumed by t h e Customer Premises Equipment
PRN =Power consumed by t h e Remote Node∨Base Station (if t h er is one)
PRN =Power consumed by t h e terminal Unit ¿
NRN =Number of customers t h at s h are a Remote Node
NRU =Number of customers t h at s h are a Terminal Unit

In equation, includes a factor of 1.5 to account for additional overheads such as external
power supplies, (electricity, distribution losses, cooling requirements), Values of access
network parameters used.

Table. 1 Values of Access Networks parameters used [6]

The next Figure is a graph of the per-customer power consumption of each access
technology as a function of the access rate.

figure 6. Power consumption vs Acces Rate [6]

[6] Jayant Baliga, University of Melbourne and National ICT Australia Robert Ayre, Kerry Hinton, and Rodney S. Tucker, University of
Melbourne “Energy Consumption in Wired and Wireless Access Networks” IEEE Communications Magazine • June 2011 71

In terms of consumption per user on these networks we can deduce several conclusions:
-A low capacities (Acces Rate<1 Mbps), most of HFC cable networks, PON, DSL) have a
similar energy consumption, to determine the energy consumption in the Access network, it
is essential to take into account the use of modem or ONU.

-A low capacities, wireless networks are more efficient in terms of consume. A grows the
demand for capacity, must devote more resources to satisfy users, growing energy
consumption exponentially.

-For high traffic volumes is that the type of network is more energy efficient PON or use
techniques in customer network equipment as “Automated sleep modems” (energy efficiency
in device end user) for offer the same access rate that PON with other technologies.
Assuming the Internet is used on average 8 h/day, automated sleep modes in customer
premises equipment could reduce the energy consumption of the access network by up to 40
percent.

- Power consumption could be realized by fast/micro sleep modes, where customer premises
equipment enters a sleep mode during periods of inactivity that are shorter than a second,
reducing the processing power.

- In access rate of 1 Mb/s, all wired access technologies are significantly underutilized. At
these low access rates, WIMAX and UMTS can flexibly share capacity among a very large
number of users, and thus can achieve high efficiency and utilization,But when we increasing
the access rate more than 1 to 10 Mb/s increases the power consumption of WiMAX and
UMTS by a factor of 2 and 4, respectively, as fewer customers can use a given radio channel
or base station and more resources must be provisioned to deliver the service.

Currently there are techniques to reduce energy consumption in existing access networks,
both wireless networks access as in fixed network access, very few implemented by service
provider.for example without changing any infrastructure could be deleted ADSL + and
ADSL of the service portafolio and only offer VDSL that allows faster transfer rates and is 70
percent more efficient in terms of energy consumption that adsl and adsl +. but other
techniques are used by the service provider for the energy consumption reduction in existing
access networks and are mostly implemented in the edge equipment. some of them are:

- Detect Inactive Link: Automatically adjust power usage according to de number of enable
users with devices connected, The service provider can maintain the ports on wired networks
with minimum energy consumption when users are not using resources

- Wake on Demand: Automatically detects the network usage to turn to the power saving
mode and wake up to the network activities such as internet access or calls

Adjust the power usage according to the data demand

- Wireless scheduling: automatically reduce power usage and radiation when a wireless
network is not in use (based on user defines scheduling)

- Smart fan in diferent devices in the access network: automatically adjust speed of fans
according to the device temperature

- Intelligent Poe: automatically adjust power usage according to the required power of each
connected device that needs power
- Detect cable lenght: automatically adjust power usage acording to the lenght of the cable
that connects a device to the central office

Now we can see the predictions about energy consumption in the future to get a general idea
will observe the following figure:

figure 7. Expected Power Consumption Access networks [6]

As already indicated in the access networks, the higher energy consumption occurs in
interconnection equipment, but as these equipment are composed of several components
(electronic, optical, power amplification, and AC/DC,power conversion). And today there
great strides in the development of components and materials that will make the more
energy efficient equipment.

As a reference the power consumption of HFC and UMTS decrease only 50% because the
most of power consumption in these access networks is in power amplifiers, it has limited
scope to improve in the future. The results in Figure. 7 indicate that the per-user power
consumption of high-speed access technologies ( PtP, FTTN,PON and WiMAX) should fall
by around 70 percent to 2020. In Wireless technologies will continue to consume at least 10
times more power than wired technologies when providing comparable access rates and
traffic volumes. Passive optical network will continue to be the most energy-efficient access
technology.

5. Network Efficiency In Core Networks

Core networks are the network level where all traffic pertaining to end users is aggregated
and distributed to their destinations. It consists on powerful routers belonging to the major
operators that interconnect different autonomous systems. It comprises a small number of it
(compared to the number of access routers), which perform all the necessary routing and
play the role of routers gateways to other operators. Its structure is high meshing between
routers belonging to the same organization and few links to network routers from other
operators. These links are usually WDM with capacities of 10 Gbps or even 40 Gbps SDH.

Througputh provided is very high (of the order of Tbps) and they are highly efficient,
energetically talking.

figure 8.Power consumption[7]

As it can be seen, as it increases the capacity of the router, it  increases their energy
efficiency. In core routers with high performance level, it is by the order of 10 nJ / bit, while
access routers (ie domestic ADSL routers), the consumption is of the order of 10 uJ / bit, 3
orders higher.

As for his participation in energy consumption, numerous studies show that while the access
speeds is low, the access network contributes over 90% of the total consumption of the
network, it  as it increases, the core network brings about a 30% of this consumption.

In this figure it is observed how as the speed of access to the network increases,the
contribution of transport and routing to consumption increases as well. Thus, as it is not
desirable to stop the advance of logical network performance and developing of new traffic-
intensive applications (IPTV, VoD) and even user behaviors that make intensive use of the
network capacity, ( not considering legal issues (p2p)), core networks will become the key
issue in the energy efficiency of telecommunication networks.

[7] Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for Ultra Broadband
Information Networks (CUBIN) University of Melbourne
figure 9. Peak access rate[8]

In this sense, it is required a multi-disciplinary approach to improve the energy efficiency of

the core network:

Some procedures studied for this purpose are:

-Improvements in technology and photonic electronic equipment.

-Routing Techniques for low consumption.
.Selectively-off network elements.
-New energy efficient protocols.
-All-Optical Network

-Improvements in technology and photonic electronic equipment.

As already said, the core network is comprised of few routers that may be geographically far
apart. The links between these routers can often be large submarine cables covering
distances of thousands of kilometers.

Despite the low consumption of optical equipment there is one aspect which can clearly
improve. Optical fibers undergo a series of effects that make at a certain distance, the signal
must be regenerated in order to remain correct. This distance is called optical reach, the
distance an optical signal can travel without being regenerated. As bigger optical reach can
be, it results in less regenerating equipment (amplifiers, etc ) and therefore less power is
required.

-Routing Techniques for low consumption.

In the current Internet, the number of hops a packet has to do given across the different
routers can be seen in this graph:

[8]Energy Consumption in Optical IP Networks Jayant Baliga, Robert Ayre, Kerry Hinton, Wayne V. Sorin, Fellow, IEEE,
and Rodney S. Tucker, Fellow, IEEE, OSA-JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 27, NO. 13, JULY 1,
2009
figure 10.Number of hops [9]

A packet, mostly, takes between 10 to 15 hops to reach its destination. Therefore, one way
to reduce consumption would be reduce this number of hops. In this way, they may reduce
the capacity requirements in the core routers.

Another improvement to achieve low consumption is called Optical Bypass, as seen in this
figure:

figure 11.Optical Bypass [10]

One process energetically costlier is optical-electrical-optical conversion (OEO). Cross-

connects and optical add-drop multiplexers route wavelengths (optical channels) avoiding
this conversion. If the common traffic could be grouped in the same wavelength , traffic could
be routed directly to the destination node without being processed in each intermediate
router.This will  drastically low the consumption.

[9]Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for Ultra Broadband
Information Networks (CUBIN) University of Melbourne
[10]Green multicast grooming based on optical bypass technology- Hou, Weigang, Guo, Lei, Cao, Jiannong, Wu, Jingjing,
Hao, Liyuan- Optical Fiber Technology, vol 17
As example, a cross-connect Calient signature wastes around two orders of magnitude less
than one standard-core router of a similar capability, in this way, one can express the lower
limit of hops as 0.5. This figure shows that reducing consumption could be achieved with this
measures.

figure 12. Power consumption decrese hops[9]

-Selectively-off network elements.

One approach to save energy is selectively power-off network elements when the traffic load
decreases, for example at night.

Shutdown means not remove the item, but maintaining an activity level in order to support
the residual traffic.

The Core Network is a network mesh enough to withstand losing any of its nodes when
traffic conditions permit, without diminishing network performance.

Obviously this is not a trivial process, and control procedures should be implemented to
optimize process closely.

-New energy efficient protocols.

It has been shown that the packet size has an influence on the routers consumption.
Therefore , the packet size can be optimized in order to minimize consumption.

Alternatively, the network can be logically segmented into two parallel subnetworks, highly
integrated, so that each carry different types of traffic according to their QoS requirements.
This type of structure reduces energy consumption.

Future energy efficient routing protocols will route traffic and save energy according to
temporal patterns, depending on the time or even the season. Even studies have been
performed such as OSPF adapt temporary traffic patterns.
-All-Optical-Network.

Some studies express the idea that switching technology in the optical domain (Optical
Packet Switching and Optical Burst Switching) offer an improvement in terms of
consumption. As we have seen, the core routers have a consumption of about 10 nJ / bit, so
to compete with future generations of these routers, optical switches should consume about
1 nJ / bit.

This alternative is quite far in time. The optical processing is an emergent technologie that by
the moment doesn’t allow to manage individually packets in the optical domain. Nowadays,
we only can switch wavelengths (10 Gbps), that is very inneficient.
.
To finish this chapter we can deduce some tips that will allow us to deduce the final
conclusions:

-The use of core networks becomes more important as increasing access speeds.
-The future trend is toward more intensive use of networks.
-The Core network equipment is more efficient than the access networks.
-Decreasing the number of hops a packet must do could lower the consumption.
-In this sense, moving to optical switching technologies will be a great advance.

4.Conclusions

As we have seen, nowadays the trend in electricity consumption due to networking is

increases rapidly. The proliferation of Internet-connected devices (tablets, smart-phones,
etc.), change of use by users and the emergence of highly demanding applications
bandwidth (VoD, IPTV, etc.) mean that the prospects are applications and users more
speed-demanding.
 
These high speeds will increase the prominence of the core network, while at low speeds the
most relevant part of consumption belongs to different access networks. to reduce energy
consumption has different aspects to consider (technology, equipment, user behaivor etc)

We have also seen that this consumption is highly inefficient, and many initiatives have been
taken to improve this efficiency. These steps can be taken in all aspects of network
operation.

These initiatives pass from modifying human behavior (rational use of devices and
networks), power management on all devices, new protocols that reduce intermediate hops
between destinations to the introduction of new technologies and paradigms clearly more
efficient in power consumption (optics and optical technologies bypass).

[10]Power Consumption in Telecommunicacion Networks: Overview and Reduction Strategies- W.Vereecken et al. – IEEE
Communications Magazine, vol 49 issue 6
[11]Energy Consumption in Optical IP Networks – J.Baliga et al. – Journal of lightwave Technology, vol 27 no 13
[12]Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for Ultra Broadband
Information Networks (CUBIN) University of Melbourne
[13]A review of energy efficiency in telecommunication networks- George Koutikas, Panagiotis Demestichas – Telfor
Journal, vol2 No 1
[14]On Determining the Optimal Optical Reach – Jane M.Simmons – Journal of Lightwave Technology , vol 23 no 3 pp
1039-104
6. References

[1] Willem Vereecken, Ward Van Heddeghem, Margot Deruyck, Bart Puype, Bart Lannoo, Wout
Joseph, idier Colle, Luc Martens, and Mario Pickavet “Power Consumption in Telecommunication
Networks: Overview and Reduction Strategies” Ghent University - IBBT, Department of Information
Technology (INTEC) in Advanced Networks and Telecommunication Systems, 2008

[2] Reducing the Energy Consumption of Networked Devices. – Bruce Nordman, Ken Cristensen –
IEEE 802.3 tutorial, july 19,2005

[3] Energy-Efficient Next-Generation Networks (E2NGN)- Pulak Chowdhury-B.S. (Bangladesh

University of Engineering & Technology) 2002 M.S. (McMaster University, Canada) 2005

[4] Reducing the Energy Consumption of Networked Devices. – Bruce Nordman, Ken Cristensen –
IEEE 802.3 tutorial, july 19,2005

[5] Energy Star – www.energystar.gov

[6] Jayant Baliga, University of Melbourne and National ICT Australia Robert Ayre, Kerry Hinton,
and Rodney S. Tucker, University of Melbourne “Energy Consumption in Wired and Wireless Access
Networks” IEEE Communications Magazine • June 2011 71

[7] Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for
Ultra Broadband Information Networks (CUBIN) University of Melbourne

[8]Energy Consumption in Optical IP Networks Jayant Baliga, Robert Ayre, Kerry Hinton, Wayne V.
Sorin, Fellow, IEEE, and Rodney S. Tucker, Fellow, IEEE, OSA-JOURNAL OF LIGHTWAVE
TECHNOLOGY, VOL. 27, NO. 13, JULY 1, 2009

[9]Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for
Ultra Broadband Information Networks (CUBIN) University of Melbourne

[10]Green multicast grooming based on optical bypass technology- Hou, Weigang, Guo, Lei, Cao,
Jiannong, Wu, Jingjing, Hao, Liyuan- Optical Fiber Technology, vol 17

[11]Power Consumption in Telecommunicacion Networks: Overview and Reduction Strategies-

W.Vereecken et al. – IEEE Communications Magazine, vol 49 issue 6

[12]Energy Consumption in Optical IP Networks – J.Baliga et al. – Journal of lightwave Technology,

vol 27 no 13

[13]Modelling Energy Consumption in IP Networks – Rod Tucker – ARC Special Research Centre for
Ultra Broadband Information Networks (CUBIN) University of Melbourne

[14]A review of energy efficiency in telecommunication networks- George Koutikas, Panagiotis

Demestichas – Telfor Journal, vol2 No 1

[15]On Determining the Optimal Optical Reach – Jane M.Simmons – Journal of Lightwave
Technology , vol 23 no 3 pp 1039-104