![Survey of energy-efficient solutions in Network
Olivier Z. Zheng, Oxford Brookes University
Abstract
Since 2000, the growth of IT has increased in a radical way, to become the first carbon dioxide producer. It
is at this time that people start to think about the global warming and the bad effect of the carbon dioxide.
Everybody try to reduce his carbon dioxide footprint. Today, computers are connected to a network. So the
main actors in IT (manufacturers, organisations…) start to work on solution to decrease the Network’s
footprint.
In this paper, different non-commercial energy-efficient solutions are described: the 802.3az standard, the
EMAN framework, the energy-efficiency metrics and the MiDORi network. Each solution is focused on a
particular part of the Network (from the interface, to the global design), it has its advantages, but it has also
its disadvantages. We’ll talk more about the disadvantages of each solution.
Each solution introduced in this paper will decrease the power consumption of the network. But to have the
best results, the chosen solution will have to be implemented in the good way to have the best impact on the
network power consumption: having Core layer routers with 802.3az ports won’t save more energy than
having Access layer switches with 802.3az ports.
I. INTRODUCTION
Since the beginning of the computing, nobody really takes care of the notion of the energy-efficient.
But since that the global warming becomes a real problem for the future.
Fig 1 Evolution of the carbon dioxide emission of several industries [1]
And as we can see on the Figure 1, the Computers and Semiconductors carbon dioxide emission grow
very fast from 1990 to 2008.
So companies try to reduce their power consumption, to reduce their carbon footprint.
Olivier Zhou Zheng is with the School of Technology, Oxford Brookes University, Wheatley Campus, Wheatley, Oxford,
OX33 1HX, United Kingdom (e-mail: olivier.zheng@ieee.org)](https://image.slidesharecdn.com/litteraturereview-110115154834-phpapp02/85/Survey-of-energy-efficient-solutions-in-Network-1-320.jpg)
![Companies have understood that they can improve their reputation and they can also save money by
reducing the power consumption.
There are some certifications (Energy Star or EPEAT) made to certify that the equipment or the
component has power consumption lower than some values, has some save mode or has a high
efficient. For example, companies may use virtualization: services virtualization on one server can
avoid having several servers. They avoid business trips by using video-conferencing solutions.
With the expansion of the utilization of IP protocol, the Network’s energy consumption becomes a new
problem.
The network manufacturers understood this and try to develop some solutions: EnergyWise and the
smart grid by Cisco, D-link Green by D-link are solutions which try to answer to this problem.
Only the non-commercial solutions made to try to improve the energy consumption, and new
approaches will be introduced.
II. BACKGROUND
There are already 2 non-commercial works, which are developing by the IETF, the IEEE.
Theses 2 solutions are focused on 2 different levels in the network: the cable at the physical layer level,
the network equipment at the application layer level.
A. Approach from the IETF
In 2008, The Register and Freeform Dynamics have made an online survey. Here is the result of the
survey.
Fig 2 Result of a survey about IT energy accounting [2]
As we can see on the Figure 2, 61% of the answer was “No, IT is not accountable”.
So it shows one thing: before trying to decrease the energy consumption, it may be interesting to
quantify this consumption.
A working group has been created in September 2010 [3] in the IETF to propose a solution: the EMAN
workgroup, EMAN stands for Energy MANagement.
The main purpose of this working group is to define a framework about energy management, and to
create a standardized (Management Information Base) on useful parameters in energy management.
This working group has released 2 drafts in September 2010:](https://image.slidesharecdn.com/litteraturereview-110115154834-phpapp02/85/Survey-of-energy-efficient-solutions-in-Network-2-320.jpg)
![-‐ The draft named draft-tychon-eman-applicability-statement-00 defines the foundations of the
EMAN framework [4].
-‐ The draft named draft-parello-eman-energy-aware-mib-00 defines an energy management
MIB [5].
The first draft explained the relation with the frameworks made by some others standardization
organisations while the second draft is more concrete with the definition of the structure of the MIB,
and the MIB itself.
Theses 2 drafts of the EMAN working group is just the beginning: in the milestones of the group, 5
RFCs submissions are planned before December 2011.
B. Approach from the IEEE
In 2007, a working group has been created in the IEEE to propose another solution: the 802.3az
workgroup. They are trying to give the ability to change the power consumption depending of the load
on the Ethernet cable.
The standard 802.3az or Energy Efficient Ethernet will allow the dynamic state transition of the energy
consumption without interruption: if there is an energy transition during a file transfer, it is transparent
for the user because there is no network interruption.
Fig 3 Illustration of the dynamic power transition with the data [6]
The standard will introduce new states in the PHY (PHYsical layer) to allow the state transition.
The standard 802.3az has just been ratified at the end of September 2010. Starting this date, there will
be the interoperability with the EEE (Energy Efficient Ethernet) function.
The main drawback of this standard is that it is a copper cable standard, there is nothing for the optical
cable. It is not a standard for this kind of cable.
III. NEW APPROACHES
There are 2 new approaches: the first approach is to calculate a metric for each network equipment to
evaluate their efficiency; the second approach is a network optimization method.
A. Energy-efficient metrics
The ECR (Energy Consumption Rating) metric, supported by the ECR initiative is a metric, which is
focus on 2 parameters: the energy consumption and the capacity of the network equipment.
The ECR is equal to the ratio of the energy consumption (in Watt) on the capacity (in bits per second).
So more the ECR metric is low, more the network equipment is energy-efficient.](https://image.slidesharecdn.com/litteraturereview-110115154834-phpapp02/85/Survey-of-energy-efficient-solutions-in-Network-3-320.jpg)
![The drawback of the ECR metric is that “ECR is a peak metric that reflects the highest performance
capacity of the device.” [7]. Because the ECR metric is only a peak metric, it may not reflect the reality
in utilization (when there are less load on the network, during the week/week-end or day/night
utilization).
To solve this problem, another metric has been defined: the EER (Energy Efficient Rating) metric.
This metric corresponds to the ratio of the capacity of the network equipment on the energy
consumption in different loads (full load, half load and idle).
The main drawback of these 2 metrics is that the metric must be used to compare equipment with the
same function: an user purchases a low ECR metric Core router and he uses it as a simple gateway
router; he will lose money because of the price and because he didn’t need a Core router for him
utilization.
Another drawback is the metric with modular equipment; empty equipment will have a higher ECR
metric than equipment with full module (even if the modules are energy consumption, the fact that they
improve the capacity is higher than the energy consumption). Even if the metric is given on his normal
utilization, the equipment may not be used in this normal utilization.
B. MiDORi (Multi- (layer, path, and resources) Dynamically Optimized Routing) Network
The MiDORi network is an optimized network with QoS (Quality of Service) requirements: all the
flows are carried on some links to optimise the energy consumption. Ports may be deactivated; nodes
may be shutdown, depending of the traffic. All the activation/deactivation processes are managed by a
PCE.
To have the optimized network starting a network topology, we need to calculate the energy
consumption of each possibility. The team of Dr Yamanaka divides the combinations in groups to
allow parallel computation with the Beeler’s algorithm on a processor [8].
In an enterprise network, this optimized network should be configured only in a part of it (the Core
layer) and should use others energy consumption solutions in the other one (the Access layer).
One drawback of this MiDORi network is that not very scalable in a non mesh topology (full mesh or
partial mesh topology), because if one link is down, one part of the network can’t communicate
anymore; it may be interesting when this part doesn’t send or receive traffic, but in others cases it’ll
become a problem.
Another drawback of the MiDORi network is that because it “aggregates the traffic and turn off vacant
links” [8], there is no more possibility to have parallel links to have load balancing: sharing the traffic
load on several links using link aggregation technology will no be possible on this network.
The MiDORi goes again his purpose with a multihoming architecture (1 network connected to Internet
using 2 links – each link connected to an ISP): in this architecture, it may have some routing policies
related to this architecture, for example, route all the VoIP on the first link and the rest on the other link
will allow the network administrator to avoid to have to define a QoS on the router with the VoIP
traffic (because it will process only VoIP traffic). On the MiDORi network, there will be no
multihoming architecture, because one link will be shutdown and all the traffic will be aggregated to
the other link; so the network administrator will have to define twice the QoS policies (once per
router).](https://image.slidesharecdn.com/litteraturereview-110115154834-phpapp02/85/Survey-of-energy-efficient-solutions-in-Network-4-320.jpg)
![The last drawback of the implementation of this network is that because all the traffic will be
aggregated on one link, the link may not be sufficient to answer to the needs, so it will create some
latency on the network, which is not really recommended in time sensitive communications.
But if the QoS is applied at the network boundaries, the QoS will be respected in this network because
it affects just the routing (and may affect the latency – see above).
IV. CONCLUSIONS
To answer to the energy consumption problem in the network, we have described 4 methods:
• The 802.3az will give the ability to set the power dynamically on the interface depending of
the traffic load (more power when there is a high load, less power during the peak off)
• The EMAN framework will give the ability to manage, to monitor and to configure the
network devices to optimize the energy consumption of the network
• The energy-efficiency metrics is more an indicator to help to choose energy-efficient
equipment, but it may be use in algorithm to determine the most energy-efficient shortest path.
• The MiDORi network is a network where the unoccupied components (links and nodes) are
deactivated to save energy, to let only the most energy-efficient shortest path activated.
Each solution may not be sufficient to save all the possible consumption, but the utilization of some of
them may give interesting results.
Each solution has to be implemented in the best position in the network to have the best impact on the
power consumption of the network.
V. REFERENCES
[1] US EAI, “U.S. Carbon Dioxide Emissions from Energy Sources 2008 Flash Estimate” 2009.
Available at http://www.eia.doe.gov/oiaf/1605/flash/flash.html [Accessed 20 November 2010]
[2] Atherton, “Should the IT department be accountable for energy use?”. The Register February 2008.
Available at http://www.theregister.co.uk/2008/02/08/it_accountable_energy_use/ [Accessed 20 November 2010]
[3] IETF, “Energy Management (eman) – Charter” 2010.
Available at https://datatracker.ietf.org/wg/eman/charter/ [Accessed 20 November 2010]
[4] Tychon E, Laherty M and Schoening B, “Energy Management (EMAN) Applicability Statement” 2010.
Available at https://datatracker.ietf.org/doc/draft-tychon-eman-applicability-statement/ [Accessed 20 November 2010]
[5] Parello J and Claise B, “Energy-aware Networks and Devices MIB” 2010.
Available at https://datatracker.ietf.org/doc/draft-parello-eman-energy-aware-mib/ [Accessed 20 November 2010]
[6] Bennett M, “Energy Efficient Ethernet – Part of Energy Efficiency Regulation” 13th July 2009.
Available at http://www.ethernetalliance.org/files/static_page_files/299ED43F-1D09-3519-
ADF31F9173B5D30F/IEEE%20Energy%20Efficiency%20and%20Regulation.pdf [Accessed 20 November 2010]
[7] Juniper, “Energy Efficiency for network equipment: two steps beyond greenwashing” 2010.
Available at http://www.juniper.net/us/en/local/pdf/whitepapers/2000284-en.pdf [Accessed 20 November 2010]
[8] Yamanaka N, Takashita H, Okamoto S and GAO S, “MiDORi: Energy efficient network based on optimizing network
design tool, remote protocol and new layer-2 switch” – Paper presented at International Conference on Optical Internet
2010, The Shilla Jeju, Korea, 11-14th July 2010.
Available at http://biblio.yamanaka.ics.keio.ac.jp/file/Yamanaka_COIN2010_WeB1-1.pdf [Accessible 20 November
2010]](https://image.slidesharecdn.com/litteraturereview-110115154834-phpapp02/85/Survey-of-energy-efficient-solutions-in-Network-5-320.jpg)
Survey of energy-efficient solutions in Network
- 1. Survey of energy-efficient solutions in Network Olivier Z. Zheng, Oxford Brookes University Abstract Since 2000, the growth of IT has increased in a radical way, to become the first carbon dioxide producer. It is at this time that people start to think about the global warming and the bad effect of the carbon dioxide. Everybody try to reduce his carbon dioxide footprint. Today, computers are connected to a network. So the main actors in IT (manufacturers, organisations…) start to work on solution to decrease the Network’s footprint. In this paper, different non-commercial energy-efficient solutions are described: the 802.3az standard, the EMAN framework, the energy-efficiency metrics and the MiDORi network. Each solution is focused on a particular part of the Network (from the interface, to the global design), it has its advantages, but it has also its disadvantages. We’ll talk more about the disadvantages of each solution. Each solution introduced in this paper will decrease the power consumption of the network. But to have the best results, the chosen solution will have to be implemented in the good way to have the best impact on the network power consumption: having Core layer routers with 802.3az ports won’t save more energy than having Access layer switches with 802.3az ports. I. INTRODUCTION Since the beginning of the computing, nobody really takes care of the notion of the energy-efficient. But since that the global warming becomes a real problem for the future. Fig 1 Evolution of the carbon dioxide emission of several industries [1] And as we can see on the Figure 1, the Computers and Semiconductors carbon dioxide emission grow very fast from 1990 to 2008. So companies try to reduce their power consumption, to reduce their carbon footprint. Olivier Zhou Zheng is with the School of Technology, Oxford Brookes University, Wheatley Campus, Wheatley, Oxford, OX33 1HX, United Kingdom (e-mail: olivier.zheng@ieee.org)
- 2. Companies have understood that they can improve their reputation and they can also save money by reducing the power consumption. There are some certifications (Energy Star or EPEAT) made to certify that the equipment or the component has power consumption lower than some values, has some save mode or has a high efficient. For example, companies may use virtualization: services virtualization on one server can avoid having several servers. They avoid business trips by using video-conferencing solutions. With the expansion of the utilization of IP protocol, the Network’s energy consumption becomes a new problem. The network manufacturers understood this and try to develop some solutions: EnergyWise and the smart grid by Cisco, D-link Green by D-link are solutions which try to answer to this problem. Only the non-commercial solutions made to try to improve the energy consumption, and new approaches will be introduced. II. BACKGROUND There are already 2 non-commercial works, which are developing by the IETF, the IEEE. Theses 2 solutions are focused on 2 different levels in the network: the cable at the physical layer level, the network equipment at the application layer level. A. Approach from the IETF In 2008, The Register and Freeform Dynamics have made an online survey. Here is the result of the survey. Fig 2 Result of a survey about IT energy accounting [2] As we can see on the Figure 2, 61% of the answer was “No, IT is not accountable”. So it shows one thing: before trying to decrease the energy consumption, it may be interesting to quantify this consumption. A working group has been created in September 2010 [3] in the IETF to propose a solution: the EMAN workgroup, EMAN stands for Energy MANagement. The main purpose of this working group is to define a framework about energy management, and to create a standardized (Management Information Base) on useful parameters in energy management. This working group has released 2 drafts in September 2010:
- 3. -‐ The draft named draft-tychon-eman-applicability-statement-00 defines the foundations of the EMAN framework [4]. -‐ The draft named draft-parello-eman-energy-aware-mib-00 defines an energy management MIB [5]. The first draft explained the relation with the frameworks made by some others standardization organisations while the second draft is more concrete with the definition of the structure of the MIB, and the MIB itself. Theses 2 drafts of the EMAN working group is just the beginning: in the milestones of the group, 5 RFCs submissions are planned before December 2011. B. Approach from the IEEE In 2007, a working group has been created in the IEEE to propose another solution: the 802.3az workgroup. They are trying to give the ability to change the power consumption depending of the load on the Ethernet cable. The standard 802.3az or Energy Efficient Ethernet will allow the dynamic state transition of the energy consumption without interruption: if there is an energy transition during a file transfer, it is transparent for the user because there is no network interruption. Fig 3 Illustration of the dynamic power transition with the data [6] The standard will introduce new states in the PHY (PHYsical layer) to allow the state transition. The standard 802.3az has just been ratified at the end of September 2010. Starting this date, there will be the interoperability with the EEE (Energy Efficient Ethernet) function. The main drawback of this standard is that it is a copper cable standard, there is nothing for the optical cable. It is not a standard for this kind of cable. III. NEW APPROACHES There are 2 new approaches: the first approach is to calculate a metric for each network equipment to evaluate their efficiency; the second approach is a network optimization method. A. Energy-efficient metrics The ECR (Energy Consumption Rating) metric, supported by the ECR initiative is a metric, which is focus on 2 parameters: the energy consumption and the capacity of the network equipment. The ECR is equal to the ratio of the energy consumption (in Watt) on the capacity (in bits per second). So more the ECR metric is low, more the network equipment is energy-efficient.
- 4. The drawback of the ECR metric is that “ECR is a peak metric that reflects the highest performance capacity of the device.” [7]. Because the ECR metric is only a peak metric, it may not reflect the reality in utilization (when there are less load on the network, during the week/week-end or day/night utilization). To solve this problem, another metric has been defined: the EER (Energy Efficient Rating) metric. This metric corresponds to the ratio of the capacity of the network equipment on the energy consumption in different loads (full load, half load and idle). The main drawback of these 2 metrics is that the metric must be used to compare equipment with the same function: an user purchases a low ECR metric Core router and he uses it as a simple gateway router; he will lose money because of the price and because he didn’t need a Core router for him utilization. Another drawback is the metric with modular equipment; empty equipment will have a higher ECR metric than equipment with full module (even if the modules are energy consumption, the fact that they improve the capacity is higher than the energy consumption). Even if the metric is given on his normal utilization, the equipment may not be used in this normal utilization. B. MiDORi (Multi- (layer, path, and resources) Dynamically Optimized Routing) Network The MiDORi network is an optimized network with QoS (Quality of Service) requirements: all the flows are carried on some links to optimise the energy consumption. Ports may be deactivated; nodes may be shutdown, depending of the traffic. All the activation/deactivation processes are managed by a PCE. To have the optimized network starting a network topology, we need to calculate the energy consumption of each possibility. The team of Dr Yamanaka divides the combinations in groups to allow parallel computation with the Beeler’s algorithm on a processor [8]. In an enterprise network, this optimized network should be configured only in a part of it (the Core layer) and should use others energy consumption solutions in the other one (the Access layer). One drawback of this MiDORi network is that not very scalable in a non mesh topology (full mesh or partial mesh topology), because if one link is down, one part of the network can’t communicate anymore; it may be interesting when this part doesn’t send or receive traffic, but in others cases it’ll become a problem. Another drawback of the MiDORi network is that because it “aggregates the traffic and turn off vacant links” [8], there is no more possibility to have parallel links to have load balancing: sharing the traffic load on several links using link aggregation technology will no be possible on this network. The MiDORi goes again his purpose with a multihoming architecture (1 network connected to Internet using 2 links – each link connected to an ISP): in this architecture, it may have some routing policies related to this architecture, for example, route all the VoIP on the first link and the rest on the other link will allow the network administrator to avoid to have to define a QoS on the router with the VoIP traffic (because it will process only VoIP traffic). On the MiDORi network, there will be no multihoming architecture, because one link will be shutdown and all the traffic will be aggregated to the other link; so the network administrator will have to define twice the QoS policies (once per router).
- 5. The last drawback of the implementation of this network is that because all the traffic will be aggregated on one link, the link may not be sufficient to answer to the needs, so it will create some latency on the network, which is not really recommended in time sensitive communications. But if the QoS is applied at the network boundaries, the QoS will be respected in this network because it affects just the routing (and may affect the latency – see above). IV. CONCLUSIONS To answer to the energy consumption problem in the network, we have described 4 methods: • The 802.3az will give the ability to set the power dynamically on the interface depending of the traffic load (more power when there is a high load, less power during the peak off) • The EMAN framework will give the ability to manage, to monitor and to configure the network devices to optimize the energy consumption of the network • The energy-efficiency metrics is more an indicator to help to choose energy-efficient equipment, but it may be use in algorithm to determine the most energy-efficient shortest path. • The MiDORi network is a network where the unoccupied components (links and nodes) are deactivated to save energy, to let only the most energy-efficient shortest path activated. Each solution may not be sufficient to save all the possible consumption, but the utilization of some of them may give interesting results. Each solution has to be implemented in the best position in the network to have the best impact on the power consumption of the network. V. REFERENCES [1] US EAI, “U.S. Carbon Dioxide Emissions from Energy Sources 2008 Flash Estimate” 2009. Available at http://www.eia.doe.gov/oiaf/1605/flash/flash.html [Accessed 20 November 2010] [2] Atherton, “Should the IT department be accountable for energy use?”. The Register February 2008. Available at http://www.theregister.co.uk/2008/02/08/it_accountable_energy_use/ [Accessed 20 November 2010] [3] IETF, “Energy Management (eman) – Charter” 2010. Available at https://datatracker.ietf.org/wg/eman/charter/ [Accessed 20 November 2010] [4] Tychon E, Laherty M and Schoening B, “Energy Management (EMAN) Applicability Statement” 2010. Available at https://datatracker.ietf.org/doc/draft-tychon-eman-applicability-statement/ [Accessed 20 November 2010] [5] Parello J and Claise B, “Energy-aware Networks and Devices MIB” 2010. Available at https://datatracker.ietf.org/doc/draft-parello-eman-energy-aware-mib/ [Accessed 20 November 2010] [6] Bennett M, “Energy Efficient Ethernet – Part of Energy Efficiency Regulation” 13th July 2009. Available at http://www.ethernetalliance.org/files/static_page_files/299ED43F-1D09-3519- ADF31F9173B5D30F/IEEE%20Energy%20Efficiency%20and%20Regulation.pdf [Accessed 20 November 2010] [7] Juniper, “Energy Efficiency for network equipment: two steps beyond greenwashing” 2010. Available at http://www.juniper.net/us/en/local/pdf/whitepapers/2000284-en.pdf [Accessed 20 November 2010] [8] Yamanaka N, Takashita H, Okamoto S and GAO S, “MiDORi: Energy efficient network based on optimizing network design tool, remote protocol and new layer-2 switch” – Paper presented at International Conference on Optical Internet 2010, The Shilla Jeju, Korea, 11-14th July 2010. Available at http://biblio.yamanaka.ics.keio.ac.jp/file/Yamanaka_COIN2010_WeB1-1.pdf [Accessible 20 November 2010]