Ray Le Maistre, TelecomTV (00:25):
Hello, you are watching the Green Network Summit, part of our year round DSP Leaders Coverage. I'm Ray Le Maistre, editorial director at TelecomTV. And today's discussion looks at the benefits of building a green radio access network or ran. Now the RAN accounts for on average about 75% of the total network energy usage for a mobile network operator. And the radio unit is the most power hungry element accounting for as much as 80% of RAN energy usage. Now the physics of radio communications means that these are far from the most efficient systems, but there are many technological and operational innovations that can help to reduce this huge power budget. We're here today to discuss how the RAN can become more energy efficient. And joining me on the program today are Saima Ansari, senior partner manager at Deutsche Telekom and Green Future Networks Lead at Industry Forum NGMN.

(01:30):
Joan, Triay Deputy Director and Network architect at Doomo Euro Labs and Rapporteur at ETSI's Industry Specification Group NFV, Francis Haysom, principal Analyst at Appledore Research and Sarat Puthenpura, chief Architect of Open Radio Access Network Initiatives at the Open Networking Foundation and the Linux Foundation's Aether Project. Hello everybody. Good to see you all. Thanks so much for taking part in this discussion, A lot to talk about. So let's get started. So the first question for today is the radio unit is responsible for significant levels of power consumption in the radio access network, but what are vendors doing to improve efficiency? So Saima, if we can start with you.

Saima Ansari, Deutsche Telekom & NGMN (02:22):
Historically speaking, with each generation of the radio networks, we have seen higher efficiencies and the supplies worse to the radio units. And actually there are several strategies that are implemented to improve the energy efficiency of that if we start from the hardware side in itself. So we will see that we have now the design of the power fusion chip sets more efficient power amplifiers and other components which bring down the overall power consumption. From the software point of view, if we see that radio units can dynamically adjust their power output based on real-time traffic demands, for example, during peak hours, the transmit power and the active component can be scaled down while still maintaining the customer experience. We have also seen in the 4G and 5G different types of sleep modes. For example, deep sleep where most rfm, the waistband components of the radio units are turned off during the low traffic period or micro sleep tx. This is another example to turn off the transmission components for a very, very short duration like somewhere in the milliseconds or something.

(03:42):
When we go, besides hardware design aspects or the software efficiency, we can also look into let's say things like white band technology usage where one single, let's say single radio unit is covering multiple frequency bands leading to fewer resources and thereby saving energy. Not to forget this also means not having multiple boxes but just one box covering multiple bands, which is again leading to a further lowering of carbon footprint as well. From the architecture point of view, there is also quite some improvements that we have seen. An example there could be, let's say the integration of the radio units with active antenna units, reducing cabling losses and so on. We have actually focused on this very topic in our latest NGMN publication called a Roadmap to Energy Efficient Mobile Networks. We are addressing there are things like passive, anti antenna, energy efficiency or frontend adapt it to you for increasing energy efficiency and this publication is already available on the N GM and webpage, so feel free to check it out as well.

Ray Le Maistre, TelecomTV (04:54):
Okay, fantastic. Well good to hear that that's documented as well somewhere where people can go and get that information. So Francis, obviously there's a lot of developments ongoing, but is there more that the vendors can be doing to help with energy efficiency in the ram?

Francis Haysom, Appledore Research (05:13):
Well Ray, I think to be honest, radio unit is not our core focus. Our core focus for research, and I'll leave others to comment on the technical detail, but I think importantly we've got to realize what is happening is that in order to use the capabilities that the radio unit is doing in terms of energy saving, whether that is to sleep hibernate, turn off scale, et cetera, et cetera, is it relies on operational environment that can make those changes safely, probably automated as well. And I think our biggest challenge in this area is not so much what the radio unit can do but the processes that would allow that kind of dynamic change that can give us these options to scale up, scale down of energy usage during the day, during the year depending on energy sources for example as well. And the reality is that most CSPs today have very conservative and manual ran operational processes and in many cases they avoid change to get a stable ran after change can often take days or weeks to actually achieve.

(06:26):
And so there's a great deal of trust issues that need to be solved, solved here. So I think we need to understand that if we're going to get to utilize that new technology, it needs to go beyond just each radio unit scaling up and down. This is this whole network problem, it's all very well scaling down an RU for example, but if somebody's moving you need to be able to bring up capacity, bring down capacity in line with that movement and often that can take minutes to actually achieve. So you've got to have that operational processes that can bring things up, bring things down. And the last thing I'll say is I think we've been talking about things like tidal rans for certainly the last 10, 12 years and it's been very limited operation and I think that reflects this challenge of the operational process of allowing these newer features to be utilized in a safe and automated way.

Ray Le Maistre, TelecomTV (07:30):
Okay, fantastic. Thanks very much Francis answers. And Joan, did you want to come in with any additional comments on this question?

Joan Triay, DOCOMO Euro-Labs & ETSI (07:40):
Yes, I would like to remark about one of the points that maybe have not been emphasized. It's about the advancements that are being done in terms of the radio configurations. It's not just only about the capabilities that are being nowadays specified and implemented in terms of being capable to switch off and on the component carriers and the RF bands either selectively some of them or all of them, but also the capabilities that are being defined to change the configuration of the radio frequencies. And here for example, some of the later advancements consider making reelection of the transmission and the reception array or also being capable to change the MIO configurations or the SBB blocks for the synchronization or even changing the transmission power of the radio units. And that together with the specification of the sleep most that has already been introduced can also bring up a lot more benefits in terms of energy efficiency also potentially making savings in energy consumption.

Ray Le Maistre, TelecomTV (09:01):
Okay, fantastic. So let's move on now and look at the role of the RAN intelligent controller. How can the RAN intelligent controller or RIC help to improve energy efficiency and which of the two RICs and their associated apps are the most relevant? Sarat, let's come to you to start on this. I know this is an area where you've been doing a lot of work.

Sarat Puthenpura, ONF & Linux Foundation (09:29):
Yes, thank you Ray. So let's see, the fundamental problem that we are trying to solve here, I mean if you look at the radio access network, we are satisfying or trying to address load which are time varying and position varying on a fixed network. So you really need to make sure that the two face it to the problem one side is energy saving, other one is maintaining the QIS, the quality of salaries. So it is a delicate balance between these two. So the way that we talked about it, I think for example Saima mentioned it turning on of components or the node itself is the way to go but you need to do it in a very, very, I would say surgical manner. So when it comes to RAN intelligent controller, let's talk about the desegregated architecture itself for the O ran open ran architecture, it basically splits the node into CU dus and so DU and RU are virtual functions and R use is the physical unit and then or virtualized score and then the both near RIC and non and the SMO.

(10:55):
So this is essentially the architecture. So when it comes to using R, it does give a lot of flexibility to levers to manage the energy consumption in the rent even though it is making the architecture a little bit more complex by desegregating. And if you don't do it carefully, the energy consumption will in fact will go up because there are multiple servers and other computing and storage infrastructure involved in implementing the CU or DU as virtual functions. So you need to do it very carefully, but the good news is that we do have enough levels to address or manage the energy consumption without losing performance. So this is a study, a series of studies that are going on by various entities including in open networking foundation, link foundation and we do have a grant from NDIA, the government in the US Department of Commerce to see how the energy consumption is varying among components and an end-to-end power model as a function of the load.

(12:12):
So RIC and associated applications like Xap and xap are associated with T, which is subsequent actions and non RT RIC are associated with R apps which takes actions, control actions which is taking a second and more. So when it comes to which application more relevant, I would say at this point I would go for non artic because you can implement a significant energy savings using our apps on controlled lobes which taking more than one second. So when it come to advanced sleep modes as Saima mentioned by going into subsecond control action and associate reap would be extremely beneficial. So for I would say for significant energy savings can be achieved using our apps running on non real time R. And the good news is that with proper adaptation of nonna R and accept with appropriate interfaces to legacy network because LT network is a workhorse today for the mobile communications and we should be able to manage energy in energy consumption in LT network calls. So we have done that and we have done some experiments and we have done some practical investigations also on that one. So the short answer is that Orion architecture does give the flexibility to implement all the sleep strategies that we are out there and need to do it carefully and we can start with the loan which and reap significant savings and move on to near accepts for advanced sleep modes. That's the answer.

Ray Le Maistre, TelecomTV (14:13):
Okay, thanks Sarat and Francis, from your research of the sector, are you seeing advances in this area with the non-real time RIC?

Francis Haysom, Appledore Research (14:25):
I think the question is possibly posing a question that doesn't really exist. Both forms of RIC are actually needed for better energy consumption in the area and they have distinct roles in terms of achieving that. So I think it's important to say it's not one or the other, it's ultimately both or the ability to one, let's be frank about the non realtime RIC, it's fundamentally analytics and imposition of intent onto the network and the near realtime. RIC is really a realtime near realtime controller for the RAN network and you need both, I think there's two areas I would call out one Sara mentioned it, this is fundamentally energy consumption and QOS are always in a slight tension and if you look at today's networks, we all want energy saving but actually many of the regulatory or the customer QOS things actually tend to favor more energy utilizations in order that we don't break QoSs the non-real time RIC.

(15:38):
That ability to understand the policy of your company, the ability to impose that intent on the network and to balance being between conflicting policies is incredibly important and that's a big role for the non realtime RIC. And also importantly in terms of being able to describe what you are doing as A CSP to your regulators and to your customers many of whom want want energy efficiency as well and want to be seeing how you are trading that as well. In terms of the near real time RIC, it's incredibly important in terms of actually managing how we can use these sleep functions, these hibernation functions, these turn on turnoffs scale functions within the whole of the RAN in fact and the ability to drive that in terms of more energy efficient handovers for example and many of the other things that happen in the real time control that will be important. So both are important and they can enable us to make the transition to a more energy efficient ran.

Ray Le Maistre, TelecomTV (16:45):
Okay, great insights there Francis. Thanks so much and Joan let's come to you for your insights on the role of the RIC here.

Joan Triay, DOCOMO Euro-Labs & ETSI (16:53):
I will be very controversial here because I actually have to agree with both Francis and Sarat. Yeah I think that there is no clear winner. Probably both of them are important however, I would like to maybe highlight the aspect of the non-real time rig and I believe that the better energy efficiency of the run is not just only a concern of the actual radio units and the radio configurations, which is where something probably the near real time rig can actuate more efficiently. But it is about the whole network deployment and here we have to consider factors that it is not just only the radio unit, it's also other parts and other network functions that are being deployed potentially on virtualized infrastructure. So there are many more factors and parameters that need to be considered and this is at this scale I would say that the non-real time R can play a critical role.

(17:54):
For instance, there's not only about the proper setup of the radio configuration, it's also about how to consolidate the workloads, how to actually manage the cloud resources and how to optimize the power consumption on them. And all this telemetry and all this information that can be fed up needs to be evaluated at a global scale. And here probably the non realtime risk as I mentioned can play the critical role and one of the things that I believe that is also important to highlight is that there's not going to be a good R apps and xap unless we can also enhance the telemetry and collect as much as much data as possible from all parts of the network. It's not only about parameters of data volumes and signal interferences or reference signal powers and qualities, it's also about the actual power consumption of the equipment that is happening not only the part of the radio but also on the typical servers that are living in the sites. So altogether I believe need to be considered.

Ray Le Maistre, TelecomTV (19:09):
Okay, fantastic. And Sarat, I believe you want to come back in and add a few more comments here as well?

Sarat Puthenpura, ONF & Linux Foundation (19:16):
Yes please. Thanks Ray. So basically I just want to clarify that both nric and non-art are critical to, they're important for energy saving that play its own role. There's no contrast about it but I need to bring in the aspect of Oran adoption here. If you look at the interfaces, the non-art R is interfacing with RAN using an interface called O one and which is fairly advanced and the other one for IC is E two, which is still not mature. So if you want to get some benefits out of, if you want to get some immediate benefit energy savings and you go for one off cell complete, sorry, the node on off, you could achieve that with nonna much faster because of the maturity of the interface And also you could do some adapters to the home interface and interface that with the, I would say traditional ran through the EMS. We have done that and we have quite a bit of data to prove that's a viable option. So all I'm saying that to start getting the benefits of Oran architecture and the R, it would probably make sense to start with the non-art R and then progress weight advanced lead modes and that's a clarification I'd like to make.

Ray Le Maistre, TelecomTV (20:50):
Okay, thanks Sarat. A lot of moving parts there obviously around the RIC. So we've talked about the radio and about the RAN intelligent controller, but what about other RAN technology improvements that will have a positive impact on energy efficiency? Joan let's start with you. What else can be looked at from the mobile operator perspective to help make their networks more power efficient?

Joan Triay, DOCOMO Euro-Labs & ETSI (21:22):
Thanks. Well here I would actually a little bit follow up on the previous question and the answer and I believe that one of the potential main key drivers beyond purely the radio and the rig is about the actual D disaggregation of the run itself and the potential for the virtualization and the cloudification of it. Basically here the benefits that we can foresee is that thanks to this aggregation we can more efficiently redistribute also the network function workloads and potentially for example, consolidate the resource usage. And besides the actual composition of the network itself in its deployment, it's about the use of the new kinds of resources that we can leverage from the cloud infrastructure. And here we have yet more components which we can use to actuate and improve on power efficiency. For example, if we need to save power or energy consumption by shutting down some of the nodes on the cloud infrastructure leveraging also new CPU architectures like for example R base which are known to provide more energy efficiency in its processing and also factors like the consolidation of the workloads in specific locations can also bring benefits in terms of, for example, the systems that we need to cool down those servers.

(23:09):
For example here, new technologies on leak, liquid equipment and also more intelligent air conditioning systems can play a critical factor. So I would say that the other factor in some, the other factor beyond the radio and the rig I would say is the actual disaggregation and the virtualization of the run and here is where we can also foresee lots of improvements in the future.

Ray Le Maistre, TelecomTV (23:37):
Okay, great. Thanks Joan and Saima, let's come to you. What other technology should be considered here

Saima Ansari, Deutsche Telekom & NGMN (23:44):
Continuing from the virtualized setup that we just heard about? I think that one of the things that I want to highlight once again is the entire idea behind controlling and it being able to monitor, right? So if we are not able to monitor then get also not able to optimize, there are several tools available. I can give you maybe an example of keplan, which is a Kubernetes based energy monitoring tool to estimate the power consumption by tracking the CPU usage. And of course it's primarily on the CPU energy consumption data, but it also correlates very well with other monitoring methods like IPMI and PD data. So it provides holistic view of energy usage in the cloud native environments and I believe that this definitely helps us to understand the power allocation also dynamically and how our workload distribution looks like and then balance the energy saving with the performance.

(24:47):
If I go away a little bit and look from the site perspective and how the sites have been built so far in our networks, from an operator point of view, the most obvious choices of course switching from legacy to the latest radio technologies and I think that it's heavily explored already. One of the things that I can add to it is the modular architecture. That could be another good option because very often we also see the placement of different passive infrastructure in the indoor units and if we go from the liquid cooling site, which I just heard also previously in the session, these are good options but probably not as advanced and there is a lot of research also going in this direction, but the outdoor cooling, which is readily available in many areas could be also used. So looking into the option of modular architecture, ensuring that the placement of different movements are actually according to the environmental conditions and then balancing that with active cooling systems also could help in reducing neurotic consumption.

Ray Le Maistre, TelecomTV (25:54):
Okay, fantastic. Thank you Saima. And Francis, can you provide us with some additional technologies to be looking at here?

Francis Haysom, Appledore Research (26:04):
I think the key one, and I mentioned it in an earlier question, I think the key technology to getting this right is operational automation. It's the ability to put automation on top of this to be able to automate change, to be able to automate change in functionality, software, et cetera, software, software distribution. And increasingly to be able to automate and put in place autonomy within the rand. The Rand runs itself to a set of intents. Those will be key to getting this energy efficiency because none, unless you can do that change trust, that change repeatedly, that change, it's very difficult to see how some of this stuff will actually be taken beyond the science lab and taken into the scale and scale network. The other area I think is the whole area of digital twins. There's a huge opportunity here, particularly as we're clarifying the network to do far more simulation of the network and to be able to better scenario plan and chaos test some of these conflicts between what is QOS level and what is energy saving, saving levels, being able to anticipate those kind of changes and being able to automate and then flow that through into configuration of the network.

(27:31):
And Joan mentioned it earlier, I think it's the really important thing is truly moving to the cloud in the RAN at the moment, the reality of what we're doing in the RAN is largely still hardware box tied. A function sits on a box and is sat in as a particular place. The more we can decouple decouple that, then the more we have the ability to be much more flexible about how and where particular network loads are cited. And also importantly to be able to use the capacity of the cloud to be able to scale that and to be much more certain about being able to use CPUs and use efficient places, data centers with efficient energy input to be able to efficiently use those and be able to scale up, be able to be certain trust in being able to run CPUs compute at a much hotter, hotter level. So I think that will also be important.

Ray Le Maistre, TelecomTV (28:35):
Okay. Yeah, a much bigger picture to consider here than with the traditional radio access networks. Sarat, have you got any other examples of technology developments that would help with energy efficiency in the ran?

Sarat Puthenpura, ONF & Linux Foundation (28:52):
Yes, I'd like to add on the comments that Saima and Joan mentioned. So what we are trying to do at this point, the part of the research work that mentioned with the Rutgers University vLab East to get an end-to-end power model, which takes the power consumption of the radio unit and also the virtual network function and is supporting the Kubernetes and bare metal and also getting ground truth using PDU and what are the minimum subtle set of parameters that you need to measure from the network to get a pretty reasonably accurate power consumption model across all those things. So that work is progressing. The idea here is that once you get a power consumption model, then you can use that to optimize. So some of the techniques that potentially we could use in the server side or the infrastructure side is to reduce the voltage on the CPUs and clock speeds and also reduce or shut down some of the functions which at low loads where you really don't want to use. So these are the additional levers that you can pull when you have a good end-to-end power consumption model and how that model can represent the power consumption under various types of load than amount of load that is offered to the network.

Ray Le Maistre, TelecomTV (30:30):
Okay, great. Thanks Sarat. Now another thing to consider is what might be done to improve energy supplies and energy management systems? Do telcos need to forge closer partnerships with energy providers? Saima, maybe we can come to you first on this.

Saima Ansari, Deutsche Telekom & NGMN (30:53):
So I wouldn't say it's going to be a surprise that NRG is one of the highest priorities in our company and the fact that we as Deutsche Telecom have already committed to use a hundred percent renewable energy all across our network definitely is a challenge because we know that it involves many different strategies coming together in play, whether it's onsite renewable energy production or going through the power purchase agreement. It all depends on what is available in which geography that we're operating in. So looking at all those factors, I would say it definitely makes sense to go into closer cooperation. But beyond that we also look at, let's say the complication that arises when you are using more and more renewable energy. One of the direct complication that you can see is the volatility of renewable energy, which one of the simplest way to go into looking at a solution could be a larger scale that is storage.

(32:07):
Because if you are generating much more energy at a certain point in time than you can use, then you are definitely relying on different battery solutions to the support. That's one of the reasons why we at Deutsche Telecom, we took also this decision to install large scale battery energy storage systems. So we plan to have a total capacity of around 300 T hours, which we also plan to use for peak shaving arbitrage and to maximize the use at the end of the day to participate in the energy markets and contribute to the stability of our grid. This is one of the options which we definitely take seriously. Another concept that I see also telco is getting very much familiar with and interested in which we also see in our Green Future Networks program at N GMs, this the so-called energy flexibility. So this is the idea of all the telcos basically referring to the ability to adapt to this electricity generation and the consumption and how to balance that for this specific period of time. And this is evolving at the moment and I do not see any concrete results there, but definitely it's developing and if we go in this direction I definitely see that it'll bring us much more closer to the energy providers and there will be a stronger partner.

Ray Le Maistre, TelecomTV (33:39):
Okay, fantastic. That sounds like really important work for all parties involved, Francis, let's come to you now. Is this a natural, stronger relationship that needs to be brokered here between the network operators and the energy providers?

Francis Haysom, Appledore Research (33:58):
I think it's useful to go back to the telecom tv, great telco debate in 23 there. We were talking about a lot about energy optimization in one of the pre-sessions. It's very clear and I remember the, I think it was the energy lead at Vodafone was talking about it. There are already very, very strong relationships with the energy providers and in fact the biggest lesson from that thing that we took away as Apple door research was that there's a huge amount of emphasis in terms of the relationship, which is really about energy arbitrage. How do I get the best price? How do I get the most efficient price? I think the piece that needs to be brought forward beyond just energy arbitrage is how do we bring this much more technical features like the ability to turn down capacity, et cetera, and bring that to the energy arbitrage, bring that in terms of the energy sources and bring that together.

(35:02):
And that was the thing for us just over a year ago was they seem to exist in very different rooms. So if we're going to make that closer partnership, it's probably getting that closer partnership between the energy suppliers, the energy buyers within telcos and the technology behind it to link into that. I think there are some signs of it. We've seen some examples with lya and Vodafone in terms of making greater ties to green energy battery usage, for example, Elisa. But I think that the key partnership is about bringing the energy procurement part of the telco in line with the technology that could enable better optimization.

Ray Le Maistre, TelecomTV (35:51):
Okay, yeah, great point. And just goes to show how useful the great telco debate and associated events are. Of course. Joan, let's come to you next on this point. How do you see the relationship between the telcos and the energy sector developing?

Joan Triay, DOCOMO Euro-Labs & ETSI (36:09):
Yeah, as others have mentioned, also critically important, not just only in the aspect of the procurement of the energy and the power, but actually also coordinating in between the power providers and the operators. For instance, very practical use case that we are developing is about how we can actually change our deployment of the network knowing in advance if there's going to be some shortage of power from the grid and that brings us new requirements, for example, getting notifications or events from the power providers and being able to understand where the shortages are going to happen so that we can actually redeploy or change or scale down our network in order to not put more pressure on the power consumption overall. And this is also complemented for example, on some of the practical use cases of deployment of the base stations that are happening in DoCoMo right now. We call them the green base stations and those base stations are equipped with the solar parallels and also with rechargeable batteries to actually overcome this potential shortages of power and at the same time to actually contribute on generation of energy. So yeah, I would say that these are quite of good practical examples that operators and Docomomo here included are working with the power providers.

Ray Le Maistre, TelecomTV (38:02):
Okay, fantastic. Great to hear that these kind of advances are being made. Well, we're coming towards the end of our session now, but we're going to close with what might seem an almost impossible question. And that question is what overall savings are possible for network operators from becoming more energy efficient in the RAN and how might these figures be improved as open ran evolves and also the associated use of ai? So Sarat, maybe we can start with you. Is there any way of quantifying the gains that can be made here?

Sarat Puthenpura, ONF & Linux Foundation (38:46):
Yeah, so the use of AI in energy saving has been used even before the open brand, but I think it has got a lot of potential. So in the context of open brand, it does give a good framework to put a models to optimize, I would say various power saving strategies, mostly via prediction. So a commonly used prediction AI application is to determine what should be the sleeping and wake up thresholds because no two cell sites in a network is the same because they are operating on different traffic conditions, geography conditions and different topologies and configurations and stuff like that. So the traditional user of AI has been in load prediction say that, oh yeah, you're expecting the load to go down at this particular time, particular situation, you can start bringing down the components of bring the node as it is down and then you keep monitoring, there is something going on and basically you need to start waking up the components so sufficiently in advance. So waking up and sleeping, especially waking up the timeframe is extremely crucial to maintain A QOS. So AI has traditionally been using in this context for that outside and within the construct of the oral architecture. So I think the way I would see is that the architecture provides even more greater flexibility to introduce AI capabilities into various energy saving strategies. And I am hopeful that there are more and more applications around that, that directions will be created by researchers and vendors. So the answer is that I'm very optimistic about it.

Ray Le Maistre, TelecomTV (40:50):
Okay, thanks very much. Joan, is there a way we can look at this and measure or predict what kind of gains might be made in terms of ran energy efficiency?

Joan Triay, DOCOMO Euro-Labs & ETSI (41:05):
Well, yes, it's a very difficult question, but I can maybe give a few figures that we are considering at DoCoMo in our case. Well, I guess that in alignment with many of other navo operators around the world, we have our targets for what we call the green targets, basically in alignment with many others trying to become carbon neutral by 2030 as well as zero net in 2040. And to achieve that, we are considering that network efficiency and energy efficiency should be by 2030, at least tenfold so above 10 times better than it is currently. And to achieve that, we are considering that we should at least achieve a savings in energy around 30% in the radio part and up to 60% depending on the base stations and also the time of the day. So these are the figures that we are considering. And while this is also a call for action for the whole industry to try to make it a reality, I'm also aware of some of the exemplary calculations that are being done.

(42:36):
For example, in the OA community, as I mentioned before, there are different mechanisms to safe power by changing engine radio configurations and so on. I'm aware that there are exemplary calculations that for example, in installments of up to 10,000 radio units with common pattern of traffic changes throughout the day, depending on the mode that can be activated, there are potentials of saving thousands of megabit per year. And it depends also on the kind of strategy that is being followed, whether it is switching off certain RF band or actually changing the configuration of the RF band, but there are potentials to actually save thousands and thousands of megawatt per year with the current patterns of the traffic that we have. And maybe these numbers also will farther evolve and improve in the future. As I foresee that, for example, new technologies in chip sets will also bring a lot more potentials for saving in the future. So yeah, I think that there are potentials and as I mentioned, I think it's a good opportunity to also raise the importance of it and a call for action for the whole industry to save energy.

Ray Le Maistre, TelecomTV (44:10):
Absolutely a great call to action and great as well too to end us on some numbers and figures that others can look to and look to try and match or even beat as well. So thanks very much for those insights. We do need to leave it there and I'm sure we'll continue this debate during the live q and a show later. For now though, thank you all very much for taking part in our discussion and if you're watching this on day two of our Green Network summit, then please send us your questions and will answer them in our live q and a show, which starts at 4:00 PM UK time. The full schedule of programs and speakers can be found on the telecom TV website and that's where you'll also find the q and a form and our poll question for this summit. For now though, thank you very much for watching and goodbye.