In The Arena by TechArena - Efficient and Scalable Solutions at the Edge with Arm

Episode Date: March 21, 2023

TechArena host Allyson Klein chats with Arm’s Chandrasekaran about the massive opportunity for innovation at the edge of the network, need for efficiency for vRAN solutions, and his organization’s strategy to deliver disruptive innovation for providers.

Alison Klein, Welcome to the Tech Arena. My name is Alison Klein, and today I'm delighted to be joined by Panch Chandra Sekharan, Head of 5G Carrier Infrastructure segment at Arm. Welcome to the program, Panch. Panch Chandra Sekharan, Thank you, Alison. Great to be here. with an introduction. I'm so excited to have Arm on the show and you represent such an important area of growth for the entire industry. Why don't you just get started with talking about Arm and your role at the company around 5G? Great. First of all, thank you, Alison. Great to be talking to you about the industry in general and Arm in particular. To introduce myself, you generally talked about my role here in Arm. Again, Panch Chandrasekharan, I head the carrier infrastructure segment at Arm. And a lot of focus is around 5G wireless cellular infrastructure today

and extending into the telco edge. And this actually covers a couple of different vectors. So the first vector is working closely with the tier one NEPs, right? So your Ericsson's and Nokia's of the world, working through the architectural requirements for the next generation of infrastructure, cellular infrastructure. And the second vector is really around the emerging paradigm of virtualization of network or VRAN or Open open ran as it's called in the industry where it's a much more of a ecosystem play which includes the the hardware vendors the oems around that the software vendors what i call the software infrastructural plays like operating systems container as a cast players and then the stack vendors, right? So the network stack vendors and

the system integrators and so on, going all the way to the operators and hyperscalers. So these are the two vectors that I deal with. And the goal really is to provide the best compute infrastructure or compute architecture for the cellular infrastructure, which obviously, I believe, is the ARM CPU architecture. Now, you just described two use case or two use case categories, if you want to think about it, that are really at the nexus of edge computing. Why are these two areas so important? And where would you characterize the industry's progress on delivering industry standard solutions to those spaces?

So if you look at the 5G infrastructure in general, the specs are somewhat crazy, right? Compared to the previous generations, like several orders of magnitude, faster, better in terms of the throughput, in terms of latency, in terms of number of connected elements with the connected machines and so on.

And this really is the fundamental variable that enables edge applications, where the whole idea of edge is having the compute closer to where the data is generated. And for that, connectivity is a very important perspective. And so these are invariably tied together. When you look at VRAN solutions, there isn't one topic in 5G networks that I think is talked about more right now. And the amount of industry focus and customer interest in this space is acute.

Why do you think Arm is the right foundation for VRAN? And what do you see in terms of your portfolio and that broader infrastructure play in this multi-billion dollar market? Yeah, sure. That's a great question. Okay, so let's step back a little bit, right? So today, if you see the kind of systems that are deployed, greater than 90%, you can say greater than 95% of the systems that are deployed are what you call embedded RAN or what we call embedded RAN or classical RAN, right? So these are usually deployed by the tier one

network equipment providers like Ericsson, Nokia, Samsung, et cetera, where the equipment itself, the hardware and software is very closely tied together, very carefully integrated. And it's essentially a custom implementation based on the use cases that are there. And then the emerging VRAN that you just highlighted is essentially taking a whole new approach to building a cellular network, which is virtualizing a lot of the functions making it more software defined and this has several advantages right so it means like the scalability enabling multiple players to come in the ease of innovation etc but here's a gap that needs to be addressed right so let's face it custom hardware with a highly parallelized implementation and hardware tends to be a lot more performant and power efficient than a software implementation that's on the CPU.

The flip side of that is you get the scalability economies of scale with ganging applications and connectivity together and fostering innovation. But the thing that we need to close is the gap in power performance. And so depending on who you talk to, the gap can be anywhere from 2 to 5x and sometimes more in terms of the power performance between the classical implementation and the virtualized implementation. So now the question is, what is the right compute architecture to close this gap? And that's where ARM really comes into play because ARM's DNA is compute efficiency. So when you use the right CPU architecture, the gap gets closer, or you can say you can get on par with the traditional implementation, with the right implementations. And now you can essentially use all of the advantages

that you'll get with cloud-native virtualized implementation of the network. For example, NEC and NTT Docomo recently published a study that they did on 5G network core, where they compared implementation on ARM architecture versus the alternative architecture and published a 72% reduction in power for the exact same workload, right?

And this raises a lot of interest. The recent MWC, the amount of inquiries that we got across the board from operators, from equipment providers, from software vendors on like ARM based implementation of the 5G infrastructure was tremendous precisely for this reason. When you look at the VRAN solutions, obviously this takes a lot of tight collaboration with the software ecosystem.

Can you talk to me about how Arm is building a foundation for software that runs on your platform and how that has progressed with ISVs in this space? Yeah, yeah, yeah. So it's another great question, right? So obviously, you know, it's not just a hardware solution. Software stacks really form an important role, right? So we look at software in multiple tiers, right? So you have what we call the important role, right? So we look at software in multiple tiers, right? So you have what we call the foundational software, right? So your operating systems,

CAS, and so on. Then you have the stacks itself, right? So network stack going from layer to an above, and then you have the physical layer or the file that is required for the cellular infrastructure. So on the foundational side, we have made some announcements recently with Red Hat on collaboration for building sustainable 5G networks using Red Hat CAS. We're also working with Wind River on enabling the Starling X-CAS platform on ARM, and we demonstrated both of these solutions at MWC. And on the stack side, we have a very broad ecosystem of software vendors that cater to both the private cellular networks as well as the public MNO-based cellular networks. So we have like several ISVs, Radices, Capgemini, Druid, et cetera, that focuses primarily on the private cellular networks.

On the public MNOs, we have made public statements around collaborating with NEC. And there are others that we are working with that we have not made public yet. So we have that covered. And another very important thing is around the physical layer implementation. And that is around, there is this big conversation going on on the architecture of how the physical layer is implemented. Namely, there is a method called look-aside acceleration of the physical layer, and there is a inline acceleration of the physical layer. I'll pause here for you to ask other questions if you want to double-click into that.

One thing I'll just say and stop is our portfolio of inline accelerators that are arm-based has been growing tremendously over the last couple of years and we demonstrated that also recently you know you've talked about a really compelling value proposition between um performance efficiency and you know a growing base of stacks that you've optimized my one question for you while you were talking, I was just thinking about it, is what is the response from telcos? And where do you think we are in terms of that hockey stick of growth that is expected for VRAM, ORAM, whatever you want to call the RAM?

Yeah. Where are we in that? And are the telcos trusting this technology at this point to go into broad scale deployment? Yeah. Yeah. That's a great question also. Okay. So the sense I got at MWC this year was we've moved from, this is the best thing since sliced bread to a sense of realism, right? So what I mean by that is like we had so many conversations that were focused on execution and design implementation and with that comes the problems and challenges we need to address right so that tells me that things are getting a little serious. And that also turned out to be another great

thing for Arm because sustainability and power came out to, bubbled up to the top of the list of concerns for operators, right? So it used to be that, you know, there were a few different things that came up, right? Innovation, TCO, power efficiency, and so on. And now like power efficiency and sustainability is bubbled up to the top i mean you can imagine why right so it's like the geopolitics and the energy prices that shot up especially in europe it kind of pushes the power efficiency and sustainability to the top of the stack right so now we are addressing more of the problems that we need to address and more of like, how do we get this going?

And we have operators that have committed to a certain percentage of their regional deployments to be based on virtualized RAN. For example, I think Vodafone has made public statements around having 30% of their deployments be based on open RAN by 2030. That's a big percentage, right? So we do see operators, some of them are ahead of others. Some of them wait and see attitude, but I generally see momentum building around this. The opportunity that you're describing is incredible, but I would assume that Arm has to do some things

to encourage the ecosystem in the right direction. What are you investing in? And are there any programs that we should know about? Yeah, absolutely, Alison. So given that the virtualized implementation and the open RAN is an ecosystem, ARM is doing a lot of things to encourage the ecosystem to work together. So last year, we launched what's called the ARM 5G Solutions Lab, where we provide all the equipment, test

equipment, RF chambers, the whole works, so that the ecosystem can come together, innovate together, build blueprint and take the solution to closer to where operator trials can happen. And so we've seen tremendous interest from the ecosystem in general. We have several solutions that we are showcasing in the Arm Solutions Lab, and we are always looking for our ecosystem to come together and develop more solutions in the areas of private networks and public networks and cloudified networks. When you look at the broader opportunity in terms of the device edge that you described, where do you see organizations implementing and what are the target use cases that you're the most excited about?

In terms of the edge applications or within the networking itself? How the networking is being deployed to deliver those types of edge applications and where do you see trends in that space? Yeah, trends-wise, it's really, there are a couple of different trends that are emerging, right? So some of the more established operators, they are looking at augmenting the existing infrastructure, right? So there is a legacy that they have to maintain. And then there is the future, which is generally believed to be a virtualized implementation of the network. And so we see more of what's called distributed RAM, where you'll have a cloud native deployment at the cellular site itself. So that's one kind of deployment use case that we see. the greenfield implementations of the newer operators tends to be a lot more cloudified,

I would say, where you have like a ganging up of what's called the distributed unit and serving multiple radios at the same time. And here the deployment can be in either telco data centers or telco cloud, or the hyperscalers and the public cloud players are also entering into the space. And so that's an area that we are very eagerly and eagerly looking at on how that is developing. Now we have all of the major U.S. hyperscalers jumping into this, jumping into this phase of access. Now you were just at MWC and you talked to, I'm sure, a tremendous amount of the industry while you were there. What do you think are the key things that the industry needs to do in 2023 to deliver upon the vision of both of these opportunities that you've described? And what should we expect from ARM?

Yeah. So from an industry perspective, I mean, I think the message is very clear, right? So it's going to be power efficiency and sustainability. So this has implications on like what kind of systems get built, what the computer architecture is and how do you manage radio, which is bulk of the power, right? So, I mean, it's like, just to give you an idea between 4G and 5G, the power consumption is two to four X greater. And like the power of a 5G base station can be like 60, 70% more power than 4G. So power is like absolutely the number one.

So you cannot have like a one kind of a solution that kind of addresses the power question. It has to be like, take the biggest chunks and solve them. So there's obviously the compute piece of it which arm can play a significant role in and then ai will see is increasingly playing a role in managing the radios ai and algorithmic management of power will become more and more critical in the next several years fantastic well thank you so much for being on the program today it was a real pleasure i really enjoyed getting to know you and learning more about what ARM is doing in this space. And I expect to see a lot more from ARM as we head further into 2023. Thank you very much, Alison. This was fun talking to you too.