In part two of this series, Erik Ekudden, Group CTO for Ericsson, and Ronnie Vasishta, Senior Vice President at Telecom NVIDIA, discuss the potential of edge and 5G.
In this podcast, Ericsson CTO, Erik Ekudden, talks to Ronnie Vasishta, SVP Telecom NVIDIA, about the next generation of immersive user interfaces.
Erik Ekudden: Welcome Ronnie Vasishta, Senior Vice President and Head of the Telco Business Unit at NVIDIA. Welcome, Ronnie.
Ronnie Vasishta: Thank you, Erik. Nice to be here — looking forward to the conversation.
Erik Ekudden: Yes, likewise really great to be able to talk about such important things as 5G, the edge, XR/AI, and all that. But let's bring the audience back to something we did together already last year. You remember, we actually announced that we had worked together between NVIDIA on your RTX GPU, the cloud XR, and your virtualization software, Ericsson with the 5G network, end-to-end with high performance, ultra-low latency, and then with Qualcomm with their with their reference design, the XR2 5G reference design for a head-mounted display. We actually showed it end-to-end how great it would be and taking all the effort to really integrate it as one solution.
I think that's really where the industry needs to go to make these collaborations work end-to-end and create new user experiences. But that's also going to take me to my first question for you Ronnie, so this is really about also the capabilities that we are putting into the network: So, how do you see the future of the edge? And perhaps, what are the things that are different from a distributed, connected data center and the nodes that you have in that?
Ronnie Vasishta: Yeah, great question. So, you know, I think as we start to look at the network, we got to think differently. Today, we talk about the near edge, the far edge — going forward, this is going to be a combination of distributed data centers. It's going to be a data center where you need it and when you need it. And those data centers are going to have the right level of provisioned AI storage and compute, and of course, connected on a high-performance network.
Erik Ekudden: I'd like to think that would also allow you to move around with your devices, and of course, the application would move as well in a seamless way. And to your point, this is really also about leveraging even higher performance real-time infrastructure that is getting integrated with the capability accelerators and the distributed fabric of computer wherever you need it.
I think this is a great opportunity both on the consumer side, and of course, on the industrial and the enterprise use cases, having that ubiquitous compute and the resources available basically everywhere — that's really how I think we can take the next step in the ecosystem. And thinking about the ecosystem here, what updates do you see on the application side in the edge, and perhaps, also in the far edge and how do you see AI playing a big role in this? I think we may be able to go into some details also about the different industrial use cases and how that will transform the industries, perhaps even touching on health care as one of the examples.
Ronnie Vasishta: Yeah, I think it's really a great time to be in an industry like ours. I won't be the first to say — every industry is going to transform within the next 10 years. If you think about today, we're already starting to see evidence of that with smart agriculture, for instance. You know drones and robots cultivating fields and applying the right amount of fertilizer to the right points and at the right time, the weather prediction that enables that. The combination of these things being much higher productivity.
Transportation, for instance, becomes a huge AI optimization capability as well. Points of that we talked about the network and data center along the points of transportation whether it be for cargo or for people. Healthcare, you mentioned, becomes much more personal, data collection and data analysis at the point of use, where you need it at the person whether it be the biometrics of a person and the ability to then apply information to that person or to a group of people on a global scale. It's really quite dramatic how industry will transform in the next 10 years.
Erik Ekudden: Yeah, I think you said it well. It's hard to see any industry that will not really be reinvented or benefit from this transition. I think it's kind of obvious that edge computing the way we think of it as a perhaps distributed fabric, we talked about it as a network compute fabric in the 2030 perspective, and you think about a network being a cognitive network with AI built in.
All of these things will, of course, change the way that we look at the digital infrastructure itself. And that will open up for completely new use cases where we have low latency, or guaranteed latency, we have the performance under an SLA, and I think that will really create this next generation of innovation whether it's on the consumer side, as we started to talk about, or perhaps even more important on the enterprise side; the efficiency gains and the new ways of working. And talking about the industry — Industry 4.0 and the transformation that are really within that. How do you see the connection then to the edge?
Ronnie Vasishta: Yeah, so you know, we talk about Industry 4.0 as a talking point, but really, if you think about it is the Fourth Industrial Revolution. There's only been three since the 1700s, so the impact is profound. And really, it's the digitization of automation, and as you start to look at as what the context we're talking about, the edge, or a data center, where you need it.
It's having that ability to have the compute, the connectivity like you talked about on a fabric with low latency for instance, as and where you need it on, whether it be on the factory floor, whether it be in the field, or whether it be any other part of the transportation industry, for instance. AI will write the software that optimizes the end devices. And we talked about, you know, drones spraying fields, or drones being able to let be leveraged in delivery of packages and parcels and medicines, and coverage across the world where today there is no coverage. So, industries are going to be transformed, as we've said, in the next 10 years and those events are already starting to happen.
Erik Ekudden: Thanks for bringing that one up, Ronnie. I think this really the notion of the network compute fabric that is pervasive. The fact that you need connectivity everywhere, of course, you need it in the cities, but you also needed in rural areas. You get great examples of agriculture about transportation, logistics, and similar. And of course, you need it indoor, so 5G is really about providing that connectivity everywhere with gigabit per second speeds and this low latency and the guarantees that allows you to explore new business opportunities.
I think down time is typically extremely expensive in these industrial cases and in the enterprise space, so we need to pay very close attention to making sure that we actually provide certainty, that we provide the reliability and the resiliency when we are using the edge to run some of these applications. But if we look a little bit further out, I also see that we can look at more extreme use cases of tomorrow. I think this is really where the extreme bandwidth, perhaps even stretching today's 5G system will come in. Highest throughput, of course, increased security and the millisecond latencies that are guaranteed — this will really open up for closed-loop industrial control system automation.
This would be about industrial robotics, it will of course be about the XR that we talked about, it will also be about really advanced feedback systems, like real-time synchronous haptic feedback, and of course, also cooperative driving and beyond that. So, I think these are just examples of the combination then of the network fabric their performance together with the cognitive part of the AIPs, and of course, the edge compute needs to come together. So, I'm very, sort of, seeing a lot of opportunities here already with some of the early commercialization that we're going through. But, as we're building this network edge, or the compute network fabric, what use cases do you see then, and what are their requirements on the network since you are close to the application developers and the actual accelerators.
Ronnie Vasishta: Yeah, so the requirements on the network is a really — the good thing in the bad thing is that it differs. You know, let’s say, if you're looking at applications on a factory floor for certain types of application, you know, there’s a robot involved, the network needs to be instantaneous, the proximity needs to detect that robot and information back within a few centimeters. If you're thinking about a car that's driving autonomously on a road as it's going out of a dark tunnel into a light and making a right-hand turn — that decision needs to be made in the car. But then, if you think about things like predictive maintenance on transportation whether it be trains, or planes, or cars, you know, the sensors will be able to listen for issues, be able to optically detect issues.
A lot of that data can be uploaded on a daily basis, or on a nightly basis — it doesn't need to be instantaneous. And there, the network behaves and the requirements on the network with different. So, that orchestration of the services on top of the network are really dependent upon application developers. I think what NVIDIA, what we have tried to do is build platforms, you know, hardware platform, the orchestration platform, and then the application platform. The application platform can run on APIs that developer can say: “I need to log into the network that can provide me a millisecond of latency.” Another application developer can say: ‘I need to log into the network that I can be high throughput, high data, you know, gigabit-per-second-type throughput, and I need to do that on a perhaps, more infrequent basis.” So, this orchestration piece is going to be very, very critical.
Erik Ekudden: I cannot agree more Ronnie. I think that is really what brings it together: the network and the compute fabric, and of course, we're already starting to see how these collaborations open up for SLA-based API access so that you can get this kind of performance. I think that this is also why we see in our research that already by 2023, some 25 percent of the 5G use cases are forecasted, to some extent, smaller or larger extent, really leverage the compute, and I think that's a fantastic evolution just over the few years until 2023. So, that was a little bit about the capabilities here and the requirements on the network. If we turn more to the business potential then and what we see there, what do you see that the edge can provide from that perspective?
Ronnie Vasishta: Yeah, so you know, some reports are already putting the economic value of the combination of AI and 5G in excess of 10 trillion dollars of economic value by the year 2035. And if you start to drill down into that kind of number, you can start to look at some of its productivity improvement in businesses. Businesses become more productive as they are able to analyze more data, use more data. There is a creation of new businesses, new businesses that are again driven by their ability to now have data as and where you need it, and connectivity as and where you need it.
Across the world, there's going to be more coverage of the network, and so now, you know, in terms of rural areas, for instance, that's going to be I think closing a lot of the digital divide that exists today, and that's a big impact on whether it be farming, or education, or growth of small towns and industries and within small towns. The economic value and the economic impact of the combination of AI and 5G, which is really the data center, AI-enabled data center in the network, you know, really cannot be underestimated, and I think the number of 10 trillion dollars plus could be much greater than that. What it really says is that this is impact is very, very large.
Erik Ekudden: I think if you tied at value to some of the capabilities that we’ve talked about, of course, a low latency is what comes to mind. But it's also other capabilities that has to do with offload of the device processing, and it’s about security, the high bandwidth and it, of course, guarantees data privacy all of those things. It’s really handled by this edge, or the network compute fabric. But one example then more on the business side that is already starting to happen, I would say, when it comes to 5G deployment — it's about gaming and cloud gaming, and that community is really very much pushing us when it comes to the limit of what the network can do.
So, I'll share a few of the stats from our latest Mobility Report. As we all know, mobile gaming is today really dominated by casual gamers, but with 5G and the edge compute, we're really bringing the capabilities that we talked about now into new segments. This is really high-quality experiences, and of course, this is accessible without expensive, unique hardware. And of course, that opens up a lot of opportunities both for developers as well as for the gamers. So, if you look at more than hundred service providers around the well that have launched commercial 5G offerings more than 20 of those have announced mobile cloud gaming services. I think that's a good testament that these capabilities are starting to be explored also commercially, and these offerings are pretty advanced already today.
And these advanced gamers, they will, of course, require much more than a best-effort system. We're talking about time-critical cloud gaming use cases, we're talking about fast, multiplayer interactions, we're talking about guarantees or 20 to 30 milliseconds end-to-end network latency, and maybe 99.9 percent likelihood of reliability on both uplink and downlink being there for you while you're gaming. I think these requirements now would then be monetized, for example, in these cloud gaming offering. I think it's already taking a foothold in the market, and I think much more to come on both the consumer side as well as on the enterprise side. But turning back then little bit to how you see, how we should realize this potential across the capabilities about technology and business.
Ronnie Vasishta: Yeah, so cloud gaming is a great example you've mentioned that NVIDIA is involved already today in many of those operators delivering that cloud gaming experience over the 5G network today. As we start to see that grow, we’ll start to see obviously, as you mentioned, the business associated with that grow, and new business models created as well that don't exist today. Most of experiences in general I think are a tremendous opportunity and puts a strain on the network. That immersive experience has to be real, and in some cases, it's actually mission critical to have those immersive experiences whether you're on the factory floor or you're doing diagnostics on a piece of equipment, or a plane, for instance, or whether you're having an immersive experience as a tourist, and you're interacting with other people around the world. The network has to support these, but what it does do is it opens up those business models that we talked about earlier and unlocks the potential that now you can do a lot more by having that information or that experience right where you need it.
I think that when we start to look at the leverage of AI, it actually helps the network cope with many of these capabilities, not just in terms of the data and throughput that's required to be delivered real time, but also in the optimization and the orchestration of the network, to be able to adapt to maybe users over here that have an immersive experience, yet the network is down here, you know. When the stadiums are full, the immersive experience is much higher the requirements, but when the stadium is empty at night, you don't need to have so much of the network configured to support that, for instance.
Erik Ekudden: I think that's a great example, Ronnie, of the flexibility that the network infrastructure needs to cater for. We have technology, such as network slicing to complement what we do with the edge compute, we have capabilities when it comes to really bringing the right capabilities of the network when it's needed, as you pointed out. This is really about bringing the network closer to the user and the application, and what we are doing is really to use the technology to create better availability, stability, and of course, performance of the infrastructure to serve those needs. And this is really where it's opened up so also for all the more advanced use cases that we talked about.
I think that some of the work that we are doing now pioneering together with partners, like what we're doing between Ericsson and NVIDIA but also with many, many others in the enterprise and industrial ecosystem. It would really unleash, of course, we will find new things to new requirements that will cater for in the infrastructure, but it will also be about collaboration in an open ecosystem system. This is about standardization, this is about open APIs, this is about how we foster this collaborative culture. And we work in for, as like the 5G ACIA, the 5G AA, the AECC just to name a few that are really bringing 5G together with edge compute, together with AI at the edge. I think these capabilities now are being than commercialized based on exposure, based on open APIs, and of course, orchestration end-to-end, as you pointed out several times.
So, I think it's a good time to reach out, it's a good time to work across the ecosystem to create these opportunities but let me go back to where we started this discussion, Ronnie. We talked a little bit about the experience from last year, when we brought the split rendering or the capabilities of a high-end device, a head-mounted display to a more lightweight form factor through offloading and putting the compute on the network edge. This is really an opportunity that is becoming real right now although they will come more capabilities going forward. How do you see that we should think about this offload or device intelligence to the network?
Ronnie Vasishta: Yeah, so I think part of the benefit of having the data center where you need it, when you need it is that a data center can be, you know, a head-mounted display, it can be a phone, it can be a car. That ability to have that optimization is an offload, as you say, is going to be critical to enabling that size, weight, and power of device, the user experience of the device. There's going to be times when the network is going to be able to offload the compute-accelerated compute from a device, for instance, and that can do that because, as you mentioned, the throughput, the latency that is enabled.
There’s going to be times when the network is going to not be required in terms of the offload you know, we mentioned earlier, a car example of making an instant decision in a car that decision needs to be made in the car, but the data that could be sent up overnight then you can offload a lot of that data and the compute to the cloud. Cloud gaming is a great example of immersive experiences that we've talked about: how much of the computer you want to have on the device or on the headset versus in the cloud: deliver the SLAs per the user experience and then offload the rest in terms of maybe training of data in AI. And that's where the platforms I think that NVIDIA has been working on will help in terms of those open APIs and the ecosystem to help develop those applications.
Erik Ekudden: Yeah, I think it's a great opportunity. All the graphic stories and computation horsepower of a rack of servers this can really be made available in this lightweight devices since, of course, as we've said many times now, latency differences between executing purely on a mobile device versus at the network edge, or the network compute fabric and the data center that's close by is below our human perception. So, this is really where we open up for application developers to execute where it makes sense. Thank you so much, Ronnie, this was a great discussion to entertain, and, of course, knowing that we are in the middle of this through the partnership as well as working with the developers, the application developers, and the broader ecosystem. I think it opens up great opportunities for the whole industry. So, thanks a lot for taking the time.
Ronnie Vasishta: Thank you very much Erik. It has been a great conversation, and as you quite rightly said that a partnership like ours that are going to help drive this ecosystem and the industry forward.
Erik Ekudden: Thank you.
Erik Ekudden: So, I've now been joined by Ericsson’s Group Chief Architect, Mattias Rimbark. So, Matthias, welcome.
Mattias Rimbark: Thank you.
Erik Ekudden: We are in the middle of 5G build out around the world that come quite far in many markets, and now we are in the early phases of building out the edge. High-demanding applications and all the things we talked about now with Ronnie. So, what's your take on where we are in the industry? The ecosystem development and the maturity?
Mattias Rimbark: Yeah, I must say, I echo Ron’s perspective on the technology readiness of the industry: the technology is in place, the network is in place, to use cases, the demand for the new applications are there, and most important of all, they have the orchestration foundation for being able to manage these workloads across the globe in a coherent fashion by guaranteeing the service-level agreements.
Erik Ekudden: Yeah, and I think that's really where it comes down to. How do we make sure that these early examples that we work with leading partners in the D15 Santa Clara Innovation Labs? How they can scale for use cases around the world, global scale? And I think that's also something that together with the orchestration piece has to come together. So, what's your thought on the remaining challenges here?
Mattias Rimbark: I think the biggest challenge is finding the balance between differentiation and globalization, making the APIs available to the post majority of the applicators in a simple way in an easy-to-use network abstraction.
Erik Ekudden: I think that we are actually making good progress across the partnerships across the ecosystem to make a simple but also to focus on the most relevant one and not completely flood the market with the capabilities that are not in primetime already. So, if I would summarize your takeaway, Matthias, again you're saying that the timing is right, prime time for these capabilities. You're also saying that the application development ecosystem is ready we have to do more when it comes to the exposure part, the API part, but it's essentially ready because of the immersive experiences, the XR but, also the AI cases, and the high-demanding, industry cases.
And then, on the challenging side, where the challenges that remains, is really have to realize that this has to be a global offering. We have to orchestrate across the global infrastructure, but we also need to strike the balance between the differentiation part, and of course, the globalization of the global part. So, I think that's really great news for the industry. We have some more work to do, but we've come a very long way right now. Thank you, Matias.
Mattias Rimbark: Thanks, Erik.
Erik Ekudden: We started this session by digging deep into the future of the edge, and the challenges that it will solve. And while many of us think about augmented and virtual reality technologies as the solutions or parts of the consumer experience today, what we know is that with 5G and the edge, this will be a capability also for enterprises and industries as they digitalize. In fact, some of the examples are really mind-boggling when it comes to using a highly capable radio network or network together with the edge to deliver high-performance capabilities. This is why augmented and virtual reality will be a game changer also for industries and enterprises when they digitalize, and this would really unleash is exponential opportunities for us.
Thank you so much for watching this episode, and also, you're welcome to follow us in the continued sessions, when it comes to edge and 5G. Thank you.