Ray Le Maistre, TelecomTV (00:00:10):
But at this point, we are going to come to our final session of the day. That session is building digital infrastructure from core to edge. Now, regardless of whether a network operator has an open or traditional network strategy or whether it'll rollies on the public cloud for a lot of workloads, even a software first service provider still needs chips, modules and boxes. But what infrastructure will they need and where will it be operated? And what does the digital service provider network look like? And what does that mean for service creation and delivery? And of course, the customer as we know, is key. Now before we begin, let's play our next highlights video, after which we'll be joined on stage by our panel experts. This highlights video is from our next gen telco infra series.

Jose Antonio Martin Martinez, Telefonica (00:01:17):
We need to be more flexible, more efficient in the way we upgrade our networks. So one key tool for us is to consider CICD tools and in service software upgrade features to ensure we are automating the deployment and testing of our network, which means to simplicity. In terms of from operational perspective,

Amith Maharaj, MTN (00:01:46):
Definitely scale helps a lot. We've just gone towards a full, let's say, refresh across. So we skipped the VNF model and we went straight from PNF towards the cloud native now. So we halfway through our three year program of modernizing the core across the group, but it was done in one process of finding the right vendors for each of the

Jose Antonio Martin Martinez, Telefonica (00:02:15):
Markets. So we are working a lot on multi-cloud environment with a software that is able to be deployed in the public cloud, in the private cloud, or in hybrid environment. So we think that with this architecture, it will be easier to introduce future generations of mobile core and we will take advantage of the current investment in the core solutions.

Amith Maharaj, MTN (00:02:39):
Fortunately, there's a lot of drive towards ensuring six G is not a complete rip and replace. So functions like the NGMN, et cetera are driving the fact that it should be pretty much software driven as well. And because of the architecture, I think we have a good level of confidence that it wouldn't be too disruptive in the future evolutions of mobile.

Ray Le Maistre, TelecomTV (00:03:16):
Okay, fantastic. Let's invite our next session speakers up onto the stage. Let's give 'em a round of applause please. And for this next session, we have Francesca Serravalle from Vodafone as our co-host. So Francesca, if you'd like to come and join me here at this brand new and very exciting table that we have.

Francesca Serravalle, Vodafone UK (00:03:44):
Yes.

Ray Le Maistre, TelecomTV (00:03:46):
Nice. Okay. Hi everybody. Let's delve into the topic of building digital infrastructure from core to edge. We're going to meet our speakers in just a second, but as you know, each session has its own poll. One question, three answer choices. You can pick just one. Here is the question, will all future network platforms need to be GPU based? So please go ahead and vote on the telecom TV website. You'll find all the polls in the DSP leaders world forum section on the agenda page. And we will take a look at the voting at the end of this session. Now, as I mentioned for this session, our co-host is Francesca Vale, head of Infrastructure and Energy at Vodafone uk. Francesca, thanks very much for joining us. Again, thanks for inviting me this year. You were here last year of course. So great to have you back. Thank you. Yeah,

Francesca Serravalle, Vodafone UK (00:04:40):
Thank you for inviting me.

Ray Le Maistre, TelecomTV (00:04:41):
Okay, excellent. You are going to give a co-host address in a moment, but first let's meet our other panelists and I'm going to ask them to briefly introduce themselves starting at the very far left with Dean.

Dean Dennis, Verizon Business (00:04:54):
Hi everybody. My name's Dean Dennis. I work for Verizon. Obviously, I'm sure you're aware who Verizon are. I run the global solutions team for emea.

Ray Le Maistre, TelecomTV (00:05:04):
Okay, thank you Dean.

Vivek Chadha, Rakuten Symphony (00:05:06):
Hey everybody. Very well aware that it's us between you and what lies ahead in the evening. So we'll try and keep it a little bit interactive. That's what I've been assured by my fellow panelists. Chad, I work for a company called Rakuten Symphonia. I had the global cloud platform business. Great to be here.

Darrell Jordan-Smith, Wind River (00:05:23):
Great. Thank you everybody for giving us the opportunity today to talk to you. And again, I'm very conscious that we're between now and drinks. I'm Daryl Jordan Smith. I'm the Chief Revenue Officer for Wind River.

Ray Le Maistre, TelecomTV (00:05:36):
Okay, excellent. Thanks everybody. So Francesca, if you'd like to make your way to our DSP leaders address Lecter.

Francesca Serravalle, Vodafone UK (00:05:46):
Thank you. Good afternoon everyone. This is the last session, is it? Yeah. So hopefully we'll wake you up. I know it's always difficult, right to keep the audience engaged at the last session. So when we talk about becoming DSP, which is the main, the title, cso, DSP leaders, what do we think about? We always think about innovative business model, innovative partnership model where we think connectivity as part of broader ecosystem bundling with platform and application. And when we ask ourself what are the key enabler, what are the catalysts, the first thoughts that comes into our mind, it's service innovation. And I know we, we've been speaking about service innovation in the previous session and it's about cloudifying the experience of service provisioning through catalog driven marketplace. And we also think at teleco, a platform virtualizing cloud defined both the IT stack and the network, the network stack where basically the BSS and OSS are all cloudified and based on microservice architecture.

(00:07:01):
And the network workload is virtualized through SDN and NV architecture. But if we think what really in Japan, all this transformation, this is about high capacity, high availability, data center centric infrastructure with age-based architecture and strong interconnection capability. And that's because we need interconnection to be able to reach the ecosystem to develop synergy, to give access to the ecosystem that our customer wants. And it's also about complying with the trends of distributed architecture, distributed architecture from an application perspective, distributed data architecture, distributed AI architecture, and even the security network workload. They're all fully distributed. So to build on top of what I say, they just want to mention what Vodafone has done as a multi-year program, which you'll probably know called Red Stream Evolution, where we have created a unique and single converged backbone, which is serving the consumer, the broadband and the enterprise customers.

(00:08:19):
And that comprises of more than 200 sites span more than 55,000 kilometers of fibers. And it's as part of this investment, it's a multi-year program. We have created an age-based architecture and we have done so because we want to decentralized the services. So those edge sites, they're hosting service platform, transport platform, IP and optical platform, internet platform and content platform really enabling us to deliver low latency services but also enabling local breakout. And as part of this rate stream evolution, we have also invested heavily on SDN through hierarchical architecture. We have an optical controller, an IP controller, and a TDO transport domain orchestrator that will automate the service provision in the service assurance and the entire life cycle. And in the future, obviously we want to leverage on the open PI to make the transport network more programmable. We have also chosen some of the sites, a number of sites to deliver our telco cloud, our virtualized core.

(00:09:41):
So it's an on-premise approach. And as based of this program, we have really looked at the demand and we have created an investment roadmap that deliver megawatt. So this is when we started talking about megawatt and not kilowatt our NAVI approach. It's based on VMware fully horizontal approach, enabling a multi-vendor approach as well. And in terms of hardware, I know you were mentioning is it a white box or not? No, it's not a white box approach. We use DE and HPE now in terms of design principle of the architecture, there were three criteria. One, it's future. So we have built a 10 gig access, hundred gig aggregation and 800 gig core scalability is another thing because it's important that we maximize the investment that we make. So adding on top rather than replacing. And then finally keeping the opex flat. And we did that with two things.

(00:11:00):
One is moving from an active olo to a passive olo. The other things is through making our infrastructure green, basically we can only deliver a green network if we have a green infrastructure. So we are deploying solar panel in our core, in our core and data centers. We are simplifying the network, we are close inside, we are switching off a platform, we are removing fca. So those are really for us important criteria to deliver what I call a green, interconnected, edge based and infrastructure. And only if we do that, we can then enable that. We can become a catalyst of transformative use case and transformative customer experience.

Ray Le Maistre, TelecomTV (00:11:56):
F, thank you very much. Please come and join me back here. Okay. I mean that's a really great insight and when do you put it that way? That's a hell of a lot that is changing in the network. So I just want to come to the panelists here and see if any of them have questions for you based on what you said to get things going. If they haven't I have, but I'm sure they will do so.

Darrell Jordan-Smith, Wind River (00:12:27):
Sure. I have a question. You mentioned energy at the end of the conversation and energy is obviously a very costly thing, but it's also a very regulated thing in the context of telecommunications. Some countries have three to 5% of their grid delivering power to cellular sites and telecommunications infrastructure. And what we're seeing there is that obviously not going down with the density that's increasing, but are there things that we can do there, like you were saying solar and putting more processing there, doing things at the edge of the network versus bringing everything back to a core data center that you can see us managing that ecosystem from an energy perspective more efficiently?

Francesca Serravalle, Vodafone UK (00:13:14):
Yeah, that's true. I mean when it comes to just, you say normally there is a clear linear relation between data growth, 5G rollout and energy consumption, but we put ourself a challenge and we say no matter if we're rolling out 5G and if the data growth, we want to keep the energy consumption flat. And it was quite a big challenge and we had to do really quite few things in the radio, in the transport, in the core. Obviously the first things that we did is really try and simplify the network as much as we can. And as we move from legacy to a fully cloudified, we try to shorten that transformation. It's a transformation program which moving from a PRIs mode of operation to a future mode time, but we're trying really to accelerate it so that we don't have a double stack of it.

(00:14:07):
Legacy and network as well. We also leverage a lot on big data. So we have a team ourself, we work in a very agile way. So we are building use case on a big data platform, which it's a blueprint architecture on the cloud. And through apo, which is made of product owner, part of my team, data scientists, scrum masters and developers, we are developing a backlog of use case, one of these use cases to create ingest data, energy data from passive infrastructure and active infrastructure and then creating a heat map where we could go and take a targeted action. So there are many things that we're doing. Obviously we're also delivering intelligence in the radio part where we could change the radio resource dynamically based on user distribution and the traffic. So what we are doing, it's a lot, but it's a really big challenge to keep the energy Flo. So it's a complete

Vivek Chadha, Rakuten Symphony (00:15:10):
Ecosystem

Francesca Serravalle, Vodafone UK (00:15:12):
Driving. Yes, really end to end.

Vivek Chadha, Rakuten Symphony (00:15:15):
I have a question related to the edge that you spoke about. Edge actually means different things to different people, but in the context of Vodafone, the question I have is related to Vodafone's been an advocate of using multiple hyperscalers as well as for part of their business. You referred big data, et cetera. And now with the combination of what you've started doing for a programmable network, including edge from an operation and control standpoint, if you zoom out, what do you think the end state for managing this kind of hybrid architecture is going to look like in a couple of years? Each of them effectively has their own way of being managed and which is suboptimal in the long run.

Francesca Serravalle, Vodafone UK (00:15:54):
I think if I look on the future, to me it's not now the case, but I was saying the future, the intelligence will be distributed and into the edge. So now when we talk about what we're trying to do, actually we're trying to create centralized service orchestration platform that can really orchestrate the end-to-end the end-to-end resources and the radio resources, the natural resources and the core resources we have of course the domain, the core and the radio and the trust will have their own domain orchestrator. But eventually if you want to monetize our network, we need the service orchestrator that we will be able to dynamically provision, manage the entire network networks lies I think in the future, which include also the edge resources. I think probably in the future I would see this intelligence having been a bit more distributed because I can see we are becoming more, we are going towards a more interconnected world, right? Lots of internet platforms interconnection with content providers, interconnection with platform providers, but also we need a more seamless experience between network providers as well. And eventually I would say in the future as we go towards more of a proconsumer we consume and we provide resources to also other wholesale and network provider, I would say we would need more of a distributed intelligence, but

Vivek Chadha, Rakuten Symphony (00:17:35):
Intelligence is shifting towards the edge slowly. So you see more of an autonomous management.

Francesca Serravalle, Vodafone UK (00:17:39):
Yeah, it has to be because as you say, the complexity will become higher. We will need to have more of a distributed intelligence, not the case yet. I don't think we are there, but that will be the future.

Dean Dennis, Verizon Business (00:17:51):
Now, just to build on that point, where do you see the investment leading in terms of new revenue streams? So is it about customer experience or is it about providing a platform to deliver over the top services that we don't even thought of yet?

Francesca Serravalle, Vodafone UK (00:18:11):
Well for now, I mean I'm head of infrastructure, so I can see the big investments is really to enable a fully end virtualized network. So we are investing heavily into expand power space. It's a big investment investing. If you have an on-prem cloud approach and you have to deliver megawatt, it's not cheap, it's expensive. It's really, it's big project. And obviously that's I think within the ramit, but within my remit, so investing in infrastructure and data center expansion to be able to then deliver containerized core networks. And once we deliver the containerized core network, then we have the end-to-end networks lives, the service space architecture. Then we can also leverage on open API approach. And then having these emerging revenues from the enterprise and the different vertical. Obviously we are in a time where CapEx investment for us it's very high because if we want to leverage on the opportunity of becoming A DSP, there is a lot we are investing on DSDN, on the NV investing, on virtualizing BNG investing a lot on expanding. We are expand, we need to expand. It's the digital fabric that eventually reach out our customers. And if we want to deliver customer experience, we want to have end-to-end control of not only not the quality of services, but quality of experience overall. So I think even we're investing heavily also on not only expanding infrastructure by expanding our fiber footprint as well in a proactive way where we use geo intelligence to map radio program, fix it, program demand, and we can create a demand driven topology.

Darrell Jordan-Smith, Wind River (00:20:13):
I've got another question. Yeah, no, absolutely. So Francisca, sorry to ask you tell that. So when you were so interested, there's lots of questions. So I spent a lot of time obviously in this industry and other industries and the capital expenditure and operational expenditures are very key topics, particularly in telecommunications. And what you said earlier is that you're trying to manage that neutrally. So what you save is what you'll reinvest back into new infrastructure. Did I understand that

Francesca Serravalle, Vodafone UK (00:20:43):
Correctly? No, what I was saying in terms of, I was trying to say let's try and keep it flat, which means as a network provider, we also, as I say, we provide services, but we buy services from others aware. We have a lot of lease line and we buy call services and we offer call services. When it comes to LI line, we used to buy IT services and we moved instead from manage IT services to dark fiber or to a passive approach where we have full control on how much on the cost, but also on the scalability of our capacity management basically.

Darrell Jordan-Smith, Wind River (00:21:30):
That's great clarification, but from my perspective, do you take an annual view of that or you take a three year, five year? How do you look at that?

Francesca Serravalle, Vodafone UK (00:21:39):
We have different view. We have a one year view, a three year view, and a long-term view. So it's the way we manage our investment program. It's quite governed. It's not like one year. And we had tried to comply with what we have said the year before when we say three years. So basically it's also a way to take accountability that when you want to invest, you invest through a long-term view. It's not strategically, not tactically, because a one year view will be a tactical view, right? So that's important. Even when we say sometimes, so I'll introduce this tool because it helped me with automating this workflow. Is this strategic? Is this long term? That's the question we always ask does is it value driven or is it just a capability that a thick one box automate one thing and that's it, and then it create complicity in the IT landscape.

Darrell Jordan-Smith, Wind River (00:22:42):
Thank you.

Ray Le Maistre, TelecomTV (00:22:42):
Okay. Well I did have a question for you as well f, but I'm going to give you a break very much. I'm going to give you a break. That's all we got time for. No, not really. It's not just you answering questions here, so we'll move on to another talking point, but great introduction and great questions. Thank you. Now we're talking here about the evolution of communications network infrastructures and talking about it from the core to the edge. And in a way this might be an oversimplification, but planning and building a communications network used to be relatively straightforward because there was pretty much a network for each service, but now it's much more or moving towards a much bigger multi-service, but cloud oriented headache that could be actually adding complexity into the mix as well. Is that fair? And if it is, how do you address that complexity and the challenges that come with that, fewer silos? If you like Darl, if we just start with you,

Darrell Jordan-Smith, Wind River (00:23:52):
I didn't think you were going to go to me first. I was hoping to cheat and listen to all my colleagues, but certainly from my perspective, what we're actually seeing is observability of network-based infrastructure, layering in elements around that around some of the things that we were talking about earlier today around AI and using some tooling there both deterministically and through inference, allowing us to get better insights to what's going on. I think you touched on it, looking at how the network might use more power at certain times in certain places and optimizing for that. I think that there's a lot of opportunity for improvement there. The problem is, this is also what we discussed earlier. Each one of these systems still pumps out data in different context and there is no way of actually collating that efficiently at this moment in time without also doing other silos, which then actually layers even more complexity across everything that's going on.

(00:24:52):
In addition to that, the company that I work for at today at Wind River, we started at the edge. So everything that we've done historically has been at the far edge in aircraft, in automobiles, in devices. And we're seeing that creep into the intelligent edge IE around the base station itself into the core. And we see more complexity there because there's brownfield complexity there with all the semiconductor technology that's there and how do you collate that? And we're seeing that evolve and become quite an interesting area of opportunity for us because we're seeing increases in needs for security increases to drive more performance out of the existing infrastructure there issues around getting power to those locations is also another big issue. And then last but not least is how do you actually drive the performance that you need almost in real time in that space in order to deliver that benefit back into the infrastructure. So it is another layer of complexity that's converging. The edge is converging with the core from an infrastructure perspective.

Ray Le Maistre, TelecomTV (00:26:04):
And Dean, if I can come to you because you are looking to deliver services using these new platforms to enterprise customers, in some ways there must be a list of pros and a list of cons as the network is evolving.

Dean Dennis, Verizon Business (00:26:23):
Yep, absolutely. Well, I think that the key word is trade-offs. So the multi-cloud environment is a serious challenge for most customers specifically in terms of IT not necessarily being a strategic decision to move in that path. It's just happening organically. So instances are popping up in the different CSPs and hyperscalers and trying to control that and manage that and then create the environment where the experience of the end user is as good as it can be and consistent as it can be is very tricky. I mean, you think about the pace of change in terms of the scur, if everybody's familiar with that. So you've got frame relay and ATM private networking, what was that, 20, 25 years. Everybody knew what they were doing, everyone was fine. Moved on to MPLS again, another stretch, another couple of decades. Then we switched to SD WAN and the software defined era in sort of 20 16, 20 17 in real terms.

(00:27:29):
And we've already moved on to SSE already moved on to SSE and the challenge of keeping pace with technology change as well as the arms race in the vendor space is a big one, right? So the approach we've taken is to abstract the technical layer from the operational layer. So the ITL stack sits air gap from the technology stack because the technology stack is just impossible to keep up with. And what do customers want? Why do they want it? You could meet five customers that have five different choices in terms of technology vendor. And so to work around that and manage that carefully, the network of a service strategies is where we've gone with it.

Ray Le Maistre, TelecomTV (00:28:16):
Okay, interesting. Vivek, from your perspective, Rakuten, how are you seeing the difference between, I mean there's still a lot of legacy network around, let's face it, and operators are at a different point in moving away from that. How are you seeing the challenges as some companies move into a more cloud oriented network? And of course, Rakuten Mobile in Japan had the greenfield opportunity to build something completely new and therefore have a completely different perspective on things.

Vivek Chadha, Rakuten Symphony (00:28:57):
So before I get to that question, I slightly disagree with the first comment you made before Dar answered that it's a headache and it's an overhead, it's a natural consequence of evolution in the technology landscape today, it's cloud native and Kubernetes five years from now, 10 years from now is going to be something else. We'll be back here hopefully having a similar conversation. So this is powerful course. I guess the choice that telcos or enterprises can make is do they lead? Do they wait and watch or are they laggards? It's not necessarily going to change anytime soon. Technology just enabling you to do things hopefully simpler, faster, better, cheaper. Coming to the question that you've asked, there is a distinct, actually there's a bit of a change as well. There was a distinct difference let's say two years ago in how the we were having with operators across the globe in terms of their desire to do something in terms of transforming digital infrastructure and the kind of concerns or questions that they had in the last 24 months or so.

(00:29:57):
The conversations I've had with the CTO CXOs are changing quite dramatically. One of the reasons they're changing is because this was also covered in the first session today, I think, and Dean you also just mentioned, I think there's growing awareness within the telco exec community about the need to abstract the infrastructure layer of technology and network from the external facing service facing view that their customers get to see. I think it was Susan who probably said, nobody says, comes to our website or your website saying, well, I'm a bank and I want to buy an SD van, and that's just a manifestation. They want to connect offices together, et cetera. So I think I completely buy that, and Rakuten and Symphony started to see that recognition. We work very closely with some of our infrastructure partners like Dell and HP who are also seeing that exact same shift where for example, on things like network transformation, there is a clear shift in temperament and understanding and appetite within the telco landscape where the CTOs are now saying, regardless of O Ran or vRAN or dran, they're saying my core and edge infrastructure, which includes transport, has to get ready to do things in a more agile, limber, elastic way, and I can do an independent discussion and vendor selection and POC or pilot about what I'm going to put on top of it.

(00:31:29):
Yes, there are some constraints or rather guardrails that they keep in mind when coming up with the strategy, but they're realizing rather than try to bake the whole enchilada in one go, they're going to have to start slicing it up. And I think there was a burger up as well on the slide. So similar analogy. So I think that is actually a very welcome shift because as I guess everybody in the room agrees, telco as a community is known for being super agile and fast. No, I'm kidding or not.

(00:32:02):
This hopefully allows us to have a slightly better conversation in a year or two from now, at least within this limited construct. And going to the theme of core to edge, unlike Thelman mentioned, the Rakuten Cloud platform, which came from the robin.io days, was also born in the edge, but also more specifically, the stateful edge will become super important for the telco world because in theory, it's great to have stateless architecture and Kubernetes, but IP addresses are kind of important when you come back to life. So you can't just stick your finger up in there and say, hopefully I'll get it back again. So solving those real world problems at scale is what things like Rakuten Mobile gave us, einen nine drills have given us. We've been able to be very fortunate that we got a chance to be at the cutting edge of delivering some of this, and now we're playing very collaboratively with the likes of Intel, Dell, hp, et cetera, on the P states, C states for energy savings.

(00:33:04):
And we've already got fairly advanced stuff coming out in a few months where we've actually been able to show real world examples of double digit savings on actual workloads. So I think that is a major change that's happened, at least in my limited view of the world for the last 24 months in terms of both the focus of the operators on how they see the digital transformation for the network. Now willing to say we have to do it in layers, not the whole thing, and the fact that some of these technologies are enabling a more elastic future, which is where you want to go to.

Ray Le Maistre, TelecomTV (00:33:38):
Okay. We've got a couple of questions, commenting questions from our online audience, which I want to come to very quickly. The first one is, and I guess this is a bit of a clarification, that having this multi-service digital infrastructure provides greater opportunities to get new services to customers. Is that a fair assessment of where we're going that this makes it easier? Dean, if I can come to you and then may

Dean Dennis, Verizon Business (00:34:12):
Francesca? Yeah, I think it's a really interesting subject, right? Because the ability to be able to create new revenue streams, new services and deliver the experience at the same time to the customer piece is obviously really very much dependent on the quality of the core network and the edge plus the access. So it is an equation that needs to be balanced. I think the investment and capital investment in the access layer is really important in that discussion. So that really does open up the Pandora's box alongside and paired with the intelligence in the edge and the compute power in the edge. And the question of, well, what do you use it for? Well, clearly there are opportunities for operational efficiencies, right? So AIOps, big data ml, I mean you think about the size of the data lake that we currently sit on in terms of just pure scale of operational data day-to-day data. And what you can do with that in terms of processing that and the insights you can get from it if you've got the capability is enormous. And the final point I would make is that the capability of the Edge device itself, the handheld device with neural et cetera, is actually opening up a whole set of opportunities on which you can hook onto, but you need to have the layers in place to be able to deliver on that.

Ray Le Maistre, TelecomTV (00:35:48):
Okay, that's a great point because that actually ties in with an online audience question we've just had in, but Francesca, I just want to come back to you for a second. We also had a question about whether the shift away from the legacy silos and more towards the new cloud oriented, edge based network, is that still a little bit too slow in terms of what's needed by the service delivery part of Vodafone?

Francesca Serravalle, Vodafone UK (00:36:23):
Well, I think actually we're doing pretty good, right? Because we're being proactive in many ways. One is investing in the infrastructure, the data centers, the fibers, which is key, right? To deliver the capacity and the services of the futures than delivering on the intelligence from SDN and NAV and et cetera. But we're also moving towards a more agile ways of working as a company, as the employees as well. We are adopting a failing fast culture where we really try and innovate many business unit and we move from this POC culture into the MVP culture where two months we deliver something, we look on whether the value has been proved or not, we either productize or if not, we just move to another use case. So it's really the cultures and the ways of working that are changing. Even if you think the fact that we are also creating a unified inventory, a unified data architecture, that in itself will enable collaboration across the different business unit. So ways of working unified inventory, unified data architecture that I think it's a key enabler for driving collaboration and driving that solution model that customers want. Customers want solution, they don't want capability and that's how we are doing it.

Ray Le Maistre, TelecomTV (00:38:02):
Okay, so the human silos are being broken down at the same time as the network silos is what it sounds like? Yes. Okay. So the question we've got here from our online audience is AI has been discussed during the sessions in one form or another, but until we address network capacity, that remains a risk of having advanced AI that's going to operate way below its true potential. So for AI to deliver on its promises, the network and the infrastructure needs to be there to support that. How will telcos overcome this? Of course, that's

Francesca Serravalle, Vodafone UK (00:38:45):
Why we have invested

Ray Le Maistre, TelecomTV (00:38:48):
In

Francesca Serravalle, Vodafone UK (00:38:49):
10 gig access a hundred gigs and I think capacity key, I mean no doubt capacity in terms of fiber asset, it's really important. And what it's also really important if we're talking about then processing awareness as well within the data centers. For us, it's a different story because we work with cloud providers and where we have strategic partnership, we created data hub, we create blueprint architecture, data pipeline, ML pipeline, and then our data scientists are building all sorts of analytics for consumer, for enterprise, and for network analytics as well. Some of those analytics are related to energy management, some others are related to the way we do CapEx investment where we use AI to help us understand where we will have a bigger ROI and so on. So I think this way we built the infrastructure, the capacity, and then we leverage on the cloud services for giving us the infrastructure to build advanced AI use case.

Ray Le Maistre, TelecomTV (00:40:01):
Okay. I just want to see if anybody else on the panel wants to address that issue. Are the networks that need to shift this traffic from one place to another to enable it to be processed and the workloads to happen? Is the infrastructure there and ready to meet the needs that we're seeing now in the industry?

Vivek Chadha, Rakuten Symphony (00:40:22):
I think that's a use case driven question. To be fair, I mean the way the question was framed probably could be interpreted in many ways, but one way to look at it, especially given the edge construct that Francesca also referred to was there are several reasonable use cases which are making themselves now available where you process data the closest to the source of its origin, which is typically the edge and you need the result back at the edge as well. You don't need to backhaul that traffic at most. You might want to send the metadata back, in which case the backbone of the network isn't necessarily your choke point. It's probably the power in the compute and the storage capacity at the edge example could be near real time, real-time facial video recognition in high sensitive areas like airports or defense establishments, you don't want to go 800 miles to ask somebody if that was VEC or Francesca, by the time you get the answer back, it's probably too late.

(00:41:17):
Or as I think Dennis had given an example a couple of weeks ago from Dell that driving on your car down the road and you see a shadow and you don't want to go back a thousand miles to say was that a deer or a shadow? Because by the time you get the answer back, there's probably a slight issue. So those examples will require to the person online who asked the question, yes, they will require a certain amount of infrastructure capacity at the edge, but it's less to do with the network, it's to do with compute storage processing on the edge. But if there's larger questions around large language models and constantly ingesting them, et cetera, with big data lake, et cetera, and trying to draw insights in real time, sure, because then you're burdening quite a bit data path across the network.

Darrell Jordan-Smith, Wind River (00:42:03):
Yeah, I was going to say, I agree with you. I think it's very use case centric and I think that it's going to evolve over time. Yes, we need more network capacity because more applications and services and innovation will interact with the network. I think Dean, you said earlier you dunno what the killer application's going to be, but you know it's going to be there. So we're going to be looking for that. But as what you were saying from our experience, we provide flight control systems, we provide safety certified systems and vehicles. That's the core of our business. A lot of that processing is evolving in that edge with more compute, more power, more utilization, more efficiency there. We see that happening there around that use case because you can't necessarily guarantee an aircraft is always going to be connected to a network at every point it flies around or a vehicle going down the highway doesn't necessarily always connect.

(00:42:59):
So you want to be able to do some of that processing locally. You want to be able to collect that data whether it was a deer or not a deal deer, and actually transition and evolve that. And at the same time, the buying nature of people that travel, say in an automobile for example, when I was younger it was all about the driving experience. Now it's evolving to being what do I do in the vehicle as I go from a to be? And the bio has changed and the same with the aircraft industry, the commercial aircraft industry, the same as your experiences going through airports. That's all the changing and it's those use cases that will evolve that, but I think the industry will lean into those use cases, help solve those problems as the network evolves.

Dean Dennis, Verizon Business (00:43:42):
Okay. Yeah. Can I just add a point to that? It's a segue really, but the capacity and elasticity of the core network is what enables all of these functions. And when you think about the criticality of these applications, these use cases, you're in a vehicle, it's got to make a decision, something's happened, it's got a process, make a decision. If you're having a network upgrade, that's not going to help. So what you've got to have is an AI ops driven, proactive, forward thinking, AI driven operational model, which actually clears the problems before they happen, anticipates, bottlenecks, et cetera, et cetera. So the two things go hand in hand. You've got the AI piece on the edge and the customer experience, but you've also got the AI piece which drives the performance of the network, which keeps everything going.

Ray Le Maistre, TelecomTV (00:44:31):
Yeah, great point. Absolutely. And

Francesca Serravalle, Vodafone UK (00:44:33):
Just also

Ray Le Maistre, TelecomTV (00:44:35):
Specifically

Francesca Serravalle, Vodafone UK (00:44:35):
That the analytics I was mentioning and the relation with the cloud providers, that was to introduce cognitive capability within our own business unit. But if we're talking about use case that are offered to enterprise or consumer that requires low latency, then it's about creating synergy with distributed cloud, which we have done where we bundled our connectivity, we host distributed cloud of cloud providers into our edge, and then we can enable those critical and transformative use case that requires very low latency. Really, as I say, I mean ai, it's fully distributed as an architecture workload, so it depends on the use case as you say.

Ray Le Maistre, TelecomTV (00:45:22):
Okay. Coming to another related talking point here, but I'm going to come to the audience in a few moments, so if you've got a question for the panel, please get ready to put your hand up. But on a related note, there's obviously a lot of an awful lot of talk these days about next generation computing, accelerated computing. Our network operators hampered if they don't have a large bank of graphic processing units, GPUs for their own operations and also to be able to use those GPUs to be able to offer the services that their customers are now wanting and needing. Is this becoming table stakes? As was mentioned earlier, if you can get your hand on them right now, but is this where all operators kind of need to think about and need to go? Anybody want to start off with that one is

Dean Dennis, Verizon Business (00:46:29):
Dean, I'm more than happy to take a stab. So I think it really depends on your investment stance and what I mean by that is we've talked about cloud providers today and logically you can buy resources from cloud providers to give you that functionality that you need. Do you invest in it and bury it in the network depending on where you see the future in terms of your revenue streams, absolutely that's an option. But then you do then have to take on board the maintenance and the ongoing end of life end of support and that whole cycle life cycle piece, which can be challenging given the pace of change.

Ray Le Maistre, TelecomTV (00:47:16):
Okay. Yeah, no, absolutely. And we can see network operators going in both directions right now. Any other

Francesca Serravalle, Vodafone UK (00:47:27):
Francisca currently we see a use case or a need to invest in GP from a NATO perspective, we do see that in terms of AI use case, that could be a driver, but more a driver of partnership rather than investing ourself as I mentioned.

Darrell Jordan-Smith, Wind River (00:47:48):
Well if you're an Nvidia shareholder or an Intel shareholder, you definitely want that capacity to be bought by everybody I'm sure. But I think I agree with you Francesca, it is not there needed today and it may evolve in the future depending on the applications and the use cases and the different business models that occur depending on what the operator wants to do.

Ray Le Maistre, TelecomTV (00:48:11):
And I think so far we've seen some links to specific services based on those investments particularly where there's a need for in-country data security. That's

Darrell Jordan-Smith, Wind River (00:48:22):
Right. Sovereign cloud for example.

Ray Le Maistre, TelecomTV (00:48:24):
Yeah,

Vivek Chadha, Rakuten Symphony (00:48:25):
I think there's massive FOMO when it comes to

Ray Le Maistre, TelecomTV (00:48:29):
That's a very expensive

Vivek Chadha, Rakuten Symphony (00:48:30):
FI don't have access to GPUs. What am I going to do? What were you doing a month ago or a year ago? I mean? But yes, if you do get a real clear AI strategy with use cases, et cetera, you could probably do something on the top line in terms of monetizing and providing new services. You could drive better insights into your network, into your billing, into your product conceptualization. But unless you are the vast minority in a few years who have absolutely no access to the ability to do AI compute, whether on your own or using a hyperscaler or some of the specialist cloud providers who are coming, I don't think it's necessarily a doomsday scenario yet almost is a bit reminiscent of the evolution in silicon every five or seven years when there's a step change, when there's a glut where people just try to get their hands on it.

(00:49:28):
I will admit that the slightly deterministic nature of AI could be one of the reasons where something comes out in the industry which wasn't anticipated, and then you could be one of those people regret saying, oh, I wish I'd put in a big chunk of money on that strategy. But if it's not obvious and if you were a fiscally responsible board, I think all of them honestly would be challenged to say, do you really have a clear business case to do a massive spend like a meta or a Microsoft today on something which is not currently a clear and present existential need for your business? It could help, but is it what's stopping you from doing your day job? Probably not right now,

Ray Le Maistre, TelecomTV (00:50:17):
Especially when there's quite a few demands on what CapEx there is as well. So let's come to the audience. I know we've got a question over here. If you say who you are and your company and your question please.

Tom Burton, Cambridge Management Consulting (00:50:31):
Yep. Hi there. It's Tom Burton here from Cambridge Management Consulting. My interest in this question is from a cyber risk and cybersecurity perspective, having compute and storage in the corp has a lot of very significant advantages. Nevermind the physical security that you can wrap around it with moats and gates and walls and if need be guns. There are also economies for scale and be able to wrap some very advanced cybersecurity technologies around a few core data centers. As you move customer computer storage out to the edge to the base stations and things like that, then obviously the ability to apply those same controls on a distributed level become significantly harder. I'm interested, how are you addressing that challenge, being able to apply equivalency in the risk management and the security out on the edge and therefore able to gain the benefits without sacrificing risk and confidence?

Ray Le Maistre, TelecomTV (00:51:42):
Okay, that's an interesting question where you want to go because it's hard because that's a question really designed for the chief security officer, isn't it?

Darrell Jordan-Smith, Wind River (00:51:54):
I'm not a chief security officer, but I deal with a lot of aerospace and defense agencies in the US and abroad. It's a big concern for them. They solve it in slightly different ways in terms of secure network-based infrastructure and what they do there. Our focus is minimizing. We have some strategies associated with minimizing the surface footprint of an operating system in that edge device to minimize cyber inference and attack and also at strategies where we work with a lot of our clients to be able to move those workloads that might reside on a public cloud into a private cloud instance. And we've, we have a software element to try and solve some of those problems. But you're right, it is a huge problem. Security at the edge is going to be both our friend in terms of securing and our enemy in terms of all the risk that would occur in that space. And I don't think the industry is mature to the point that you suggest unless it's a controlled DOD type instance that we have that solved.

Vivek Chadha, Rakuten Symphony (00:52:58):
I'd just like to add something on that. So I think it's a very valid question, but the current evolution in the technology for edge, not just the compute and the storage side, but the command and control infrastructure is actually evolving quite quickly that there is a reasonable level of risk mitigated control available today. I think the question would be if it's just a few distributed ED sites, then does the cost of that control plane and the security plane justify the investment at this point? I think only if it's a scale deployment, it'll probably make sense. But for example, and this is true even for some hyperscalers who are now going to the edge because that's where the cloud is moving, you can actually have donged encryption devices which are geo centered, which means if somebody steals a rack and runs away, it blanks out the blade and it's asymmetrically encrypted, which means you can't read anything on that ever.

(00:53:52):
It's gone. If you're an operator, you can recover it because you've got backups in your cloud account. So that's one example, but it's also, an analogy might not be the best analogy, but if you go back a few years, the little bit of money we had was stored in a safe place in a vault in a bank and then came these newfangled things called ATMs, which is what's happening with the edge now. The bank is the core network and the ATMs are the edge, and they did get secured over time. There were some nice movies made about somebody trying to steal an ATM, wonder how much they got out of it. But so I think the question is valid. The attack surface does get expanded. There is a ton of compliance directives from nist, et cetera, and if you look at that body of collateral that's being applied, not just with DOD, but anything that's sensitive, whether it's CI data, PI data, customer sensitive information, et cetera, I think this problem will get reasonably well tackled in the near future.

Dean Dennis, Verizon Business (00:54:55):
And just to add to that, there's another dimension which is that not only the data in the computer has moved out from the boundaries of the physical enterprise or location, but also the users have so over covid anywhere, anytime, any place. So access to it, who has access to what, where and when. And then of course there's the regulatory challenges around that depending on where the data is, where it's being transported and what happens to it, where it reach its destination. So there's lots to consider there. So point taken. Absolutely.

Francesca Serravalle, Vodafone UK (00:55:31):
Yeah, I agree, agree. I think this is important. It's not just the fact that our network is becoming more distributed is also the fact that the application are moving to the cloud and the users are fully distributed and fully distributed. So that creates, we're moving from a perimeter type of security to something where the threat can come from anywhere and we really need a zero trust model and a security architecture. By design. I can only tell that we have a very strong security operating model and everything that I do, any change, any use case, any data that goes out, it has to go through the security operating function and they do an EM impact assessment and they can come back with yes or not and this is what need to be done. Obviously I think as we move from perimeter security to more of, I think more of holistic approach, probably peering with security infrastructure security provider where they have kind of layer one to seven type of capability that they can offer, then I think this is key as well as maybe changing and transforming your security operations center to become more data and AI driven.

Ray Le Maistre, TelecomTV (00:56:57):
Okay, great. Great question. Great answers. Thank you. I think we have another question over in the file corner again. Name, rank and serial number please.

Francis Haysom, Appledore Research (00:57:07):
Francis Haysom from Appledore Research. Again, interesting. The conversation mainly talked about technology, but one of the aspects I think that we kind of miss is the existing infrastructure, investment and investment cycle. It's extremely slow in organizations. You could say it in some senses, it makes a Soviet year, a five-year plan looks agile. But I'm just interested in do we need to look at as much less about the technology but actually our agility in terms of changing the way that we buy infrastructure or the way we procure infrastructure? Sorry, to pick on Francesca again, in the time that you've been evolving your infrastructure plan, Vodafone has divested itself of I think two major European operators. How is that changing your investment in this infrastructure plan given quite a major change in the business itself? Sorry, it's not about Vodafone, but do we need to change the speed of the cycle?

Francesca Serravalle, Vodafone UK (00:58:15):
It's a great question and something it kept me awake a few nights in those years and the reason for that that you can see the service, everything. So software, very agile, very fast, and then you actually, you need to build the infrastructure, expand the data centers, and it's a two years project or one and a half year as you say. So how do we make sure that we align the infrastructure expansion with the demand roadmap? So it's really an important thing. And there are two questions. One is we need to be faced a little bit. We have two to three years facing, right? I need to look ahead of three years because it takes one year, two years to build and expand the data centers and deliver megawatts. It's not an easy thing. The other thing question is, okay, so I invest millions now in this project, how far do I go in terms of demand?

(00:59:18):
Do I look at three, five years, 10 years, the month? How much do I need to invest? We are always thinking, oh yes, there was one question where you're saying we have now we are going towards a cloud enabled network. What challenge do we have? That is a key challenge because we always see, oh, it's about delivering a scalable infrastructure, but we still need to define the specification of all the electrical and mechanicals and everything when we expand on data centers and sub components. Once you reach the maximum capacity, you cannot add on top some other. You just need to throw away like generators and they cost a lot. So you really need to create the right compromise and say, okay, a buyer generators looking at the 10 years demand, but I buy other component maybe looking at three years demand that will the OI will be. So it's not, I think it's a great question. It's not an easy answer probably there will be different view, but I think it's a compromise in looking at what's the capacity upgrade component by component or you need to replace completely,

Darrell Jordan-Smith, Wind River (01:00:33):
I think it's about the disaggregation that gives you the flexibility, but with flexibility you get complexity. And can the industry address the complexity? It comes back to the points that we were asking ourselves earlier. So certainly from our perspective, we believe that disaggregation is a good thing. We think that gives you that level of flexibility. So you can swap out a generate or you can swap out bits of infrastructure without impacting other elements in that, and it gives you the flexibility in terms of choice and then commercially that also drives cost in the right direction in theory. So from our perspective, we think disaggregation is part of the answer, but with that it's the complexity that goes with it. And I'm pretty sure you guys have looked at that as well.

Vivek Chadha, Rakuten Symphony (01:01:19):
I mean, I agree on the complexity versus the choice part, but I have a slightly different take on the question that Francis posed and I think it'll be interesting to see Vodafone or Francesca's view on this as a thought experiment ignoring the laws of computer science and physics, assume that in five years data demand is going to go up by a hundred x in the uk. How much more network capacity are you going to seriously build, not just Vodafone, all the operators in the uk, it's a thought experiment. It might never go up by a hundred x or it could go up to that level in 20 years. And the reason I'm posing that as a starting point is that there was a reasonably valent attempt made at ran sharing several years ago in multiple parts of the world. Some of it succeeded, some of it didn't. Various reasons we had passive infrastructure sharing in the UK Beacon, CTIL, et cetera.

(01:02:12):
Maybe it's time to re-look at that model, but not for the ran bit Now, not for the passive bit, but if you have a kind of harmonious requirement for back hauling traffic from your respective endpoints and the endpoints are now mobile because your customers are no longer fixed points maybe over the coming years, that starts becoming a potential collaboration model. While the telcos still compete on everything else, there are bits of the infrastructure which they may want to look at and say, well, this makes sense for everybody. It's better for the consumer, better for the environment. I don't have to dig up roads all the time, and how much fiber are we going to lay individually? So that's one way to look at it. There is already a small precursor to this, so it's perhaps not a complete abstract notion. It hasn't made its way into the consumer telco space.

(01:03:03):
To the best of my knowledge so far, at least not directly, the neutral host model, the kind of things people like building networks, et cetera are doing. They are solving an existential need in stadiums and mass transit systems and smart cities by saying every operator wants access and the ability to provide services there. But it's financially quite stressful for every individual operator to build all this infrastructure. So they're sort of acting on behalf of the venue or the tenant owner saying, I'll take care of the infrastructure, build and act as an aggregation point. But then you have the ability now to provide service from any operator. Now, I know it's not quite the same example as a full network, but in the theme of disaggregation, to go back to your point, Francis, maybe one of the elements, and again goes back to the point of how do you eat an elephant bit by bit, maybe the answer is saying, can I solve the whole debate in one go? Probably not because it's too complex, too many moving parts. Some of it's competitive and it is differentiated competitive, so you don't want to share it with the other guy. Some of it is not. So maybe that could be something to explore in the coming years.

(01:04:16):
It has got nothing to do with technology.

Francesca Serravalle, Vodafone UK (01:04:19):
I think that will probably be the trend. And I'm sure if you go and speak to Mark Gilmore, sorry if I mention your name, but it believes this is going to be the trend, right? Sharing more on the back hole and the infrastructure part. And I believe that could be, I think, and I can support this because we are seeing more dark fiber deals coming our way. So it's a way to say, oh, let me use, you have a lot of infrastructure capacity. Can I see? So we are using commercial agreement rather than infrastructure sharing, but it's commercial agreement that I can see that happening. I think there was one more point that I wanted to make. You made this one and another one, sorry, segregation.

Vivek Chadha, Rakuten Symphony (01:05:09):
What

Ray Le Maistre, TelecomTV (01:05:09):
You need is a Negroni, I

Francesca Serravalle, Vodafone UK (01:05:11):
Think. Yes.

Vivek Chadha, Rakuten Symphony (01:05:14):
We'll coming to start shortly. Is it cloud native though?

Ray Le Maistre, TelecomTV (01:05:19):
We'll be coming to it very shortly. Actually. We're getting a bit close on time, but there was a very quick question. And it kind of almost relates to that because if you're talking about sharing access infrastructure and neutral hosts, et cetera, that then puts a greater focus on, so where do the operators compete? Well, they compete because they have separate core platforms. They have the brains of their network and their subscriber management, et cetera, is still separate. But what we've seen in recent years, and this is just how it is, of course, is massive, massive investments in 5G in particular in the mobile side, and then fiber access. So an awful lot in the access. But the 5G core, which of course is what enables a lot of service creation, delivery. I saw a stat the other day in Europe, there's very few operators in Europe with a 5G standalone core deployed. So all these amazing services, network slicing, et cetera, et cetera, they're not there yet. So is this the time where there needs to be a much greater focus on what's happening in the core, the logical core of the networks to be able to develop the services that can make use of all those access investments? I'm going to come down the line, but if we can get some relatively quick answers because we are pretty close to time. So to Dean. Yeah.

Dean Dennis, Verizon Business (01:06:51):
Yeah. Well, I mean you'll all be aware of the significant investment we've made in the US in the C-band and public and private 5G space over here in Europe. We don't have a macro network capability. So it's purely private. And to your point, I think the ability to really drive value out of these investments is hinged on the capability that's built in the core and what you can enable and what you can trigger and how you can build on that and get the flywheel effect to move away from solving tricky connectivity issues into the real value chain of the customer. So that's my view.

Ray Le Maistre, TelecomTV (01:07:38):
Okay. Vivek, are you seeing more focus now on core investments and core deployments?

Vivek Chadha, Rakuten Symphony (01:07:47):
Specifically core transformation, yes. Whether it's SA or NSA is perhaps a secondary level detail, but when it comes to 5G, there's been a very significant uptake in telcos now saying we need to sort out core first.

Ray Le Maistre, TelecomTV (01:08:08):
Okay, Daryl?

Darrell Jordan-Smith, Wind River (01:08:10):
Well, I think standalone is going to drive economic benefits, but with the economic benefits to come new services, and I think the operators are kind of stuck there. What are the new services that are going to justify that investment? And to your point, I do see a lot of operators, not necessarily in Europe unfortunately, but around other parts of the world investing more and more in standalone core

Ray Le Maistre, TelecomTV (01:08:35):
Francesca.

Francesca Serravalle, Vodafone UK (01:08:35):
Yes, we're investing equal in access transport aggregation and in the core as well. And I think it's important, right, because eventually you want to monetize the investment through networks slicing, but also to monetize the net slicing you need the service orchestrator. In the business case, our service orchestrator is linked to 5G rollout, so it's really important to invest on all the domains.

Ray Le Maistre, TelecomTV (01:09:01):
Yes, a lot of puzzle pieces to put together, as you pointed out so well in your opening address there, because that really set the scene incredibly well. So we are almost out of time for this session. That time went really, really quickly, a great discussion, but we do have to end there. But before we do, let's take a look at the poll results. And as a reminder, the question we asked, will all future network platforms need to be GPU based? Let's see what the results are. Okay. Alright. Will NVIDIA's share price go down right now as a result of this? I'm not sure. Not sure. We need to put this online, see what happens. But that's a pretty interesting, not sure is pretty high there though. So we'll see how this develops because I think as we've mentioned, these polls that are on the telecom TV website, they're not closed because we've entered this session.

(01:10:02):
These polls are open. So if you haven't voted, please do go and vote. So that's the end of session four, and that is also the close of the opening day of this year's DSP lead as well forum. We are going to be back on stage tomorrow morning at 9 25, a bit later, so you can stay up a bit later. And that's 9 25 UK time for the first session of day two, and that's focused on unlocking platform opportunities with network APIs. In the meantime, we have a final edition of extra shot over in the corner for our online viewers. For our in-person attendees here in Windsor, we have our annual evening social. It starts at 6:30 PM and will be held in the orchard room, which is near the entrance to the hotel. So see you all. For those of you here, see you all in the orchard room at six 30. But for now, let's say thank you to all of our guests. Great.