The Future of Infrastructure

Geof: Good day and welcome back to the Arm viewpoints podcast. Our episode today is the future of infrastructure and as it’s a very hot topic right now. We’re lucky enough to have a guest who is at the forefront of data infrastructure innovation, Chris Burgey, senior vice president and general manager of the infrastructure line of business at Arm. His organization is responsible for the proliferation of Arm-based solutions throughout the data infrastructure of today and tomorrow, from Cloud computing to the network edge and we’re very excited to have him here today. Welcome,

Chris: Geof it’s a pleasure to be here with you.

Geof: So Chris, maybe we can start by just exploring what is Arm’s role in infrastructure, maybe you can tell us a bit about that?

Chris: For Arm, we think about infrastructure being everything from high performance computing, which is, you know, the world’s largest supercomputers that have even played a role in creating some of these COVID vaccines that we’re starting to see distributed to traditional data centers that are really cloud-based data centers, and then to the edge and including the networking infrastructure, which has moved more and more to software defined, as well as you know, things like 5G, and obviously, what were the essential role 5G has become. You can imagine that all of all of those categories of infrastructure have a lot of compute in them and Arm has been played a role, a different role in different markets for quite some time but I think that Arm is, is quickly coming to the forefront in almost all of those markets.

Geof: So how did arm get started in it? And why? What was the impetus?

Chris: Well, so Arm has, you know, a 30-year history of building performant and very power efficient processors. And, you know, we’ve had, you know, it’s almost been a decade of actually, taking the Arm processors are quite prevalent in today’s smartphones or IoT devices, and looking at how can we scale those up into some of the more performant markets like HPC, high performance computing and data centers. So that’s really the journey that we’ve been on most recently and it’s something that we’ve gotten quite a bit of success most recently around, you know, I think AWS has been quite vocal about their use of arm and infrastructure with a first had done a Graviton processor and now they have the Graviton two processor that they’re offering. And they’ve landed some pretty impressive customers between Twitter, Formula One, Snap Lyft, and many others. But in that largely started from AWS actually buying a start-up called Annapurna back in 2016, that was basically starting to try to drive this, this idea of Arm into infrastructure or into data centers. And around the same time, we also had a company called Ampere start-up in Silicon Valley that was funded. And it started also to make some success around getting Arm and infrastructure, and today, we have partners such as Oracle and others that have announced their intention to use the ampere processor. So it’s been a long time. But we’re really excited about about the progress and back in 2018 we actually made an announcement of kind of Arm’s investment in this space, and we branded a new line of products called an Arm Neoverse. And at that time we launched our first two platforms, which was the N1 E1 of Arm Neoverse. And since then, we’ve actually announced some follow on generations, but it’s really been a long journey, as I mentioned. Yeah. And it sounds like impressive results on the journey so far.

Geof: Now, I’m going pull the lens back a bit and ask you about how you see the state of the world’s data infrastructure right now?

Chris: That’s a great question. You know, today, infrastructure is as essential as it’s ever been. And I think we’ve kind of gone especially seeing that over this last year plus where we’ve gone through COVID-19. And what did it mean to go to school from home or work from home? And what’s been amazing for people and myself included is just how well the computing infrastructure, the communications infrastructure has been able to basically keep up with that demand. And so there’s a couple interesting, kind of underlying underpinning reasons for that. One is that infrastructure as continued to become more software defined, and what the real value of it being software defined is the portability. So this idea of what type of requirements on the infrastructure can quite change quite rapidly and in fact, change in ways that we would not have really foreseen, you know, pre-COVID-19. So really being software defined and having that flexibility is definitely been key.

The second part of infrastructure computing today is how really more and more of the computing is, is moving towards trying to move it towards where the where the data is happening, which you’ll hear people talk about edge computing, or, you know, this idea of computing out at the edge of the network versus, let’s say, more of a monolithic approach. And, and that’s something that is quite near and dear to our hearts, because when you’re actually doing edge computing, things like power consumption, as well, as you know, custom compute or really compute to focus on that specific workload, are both two strong points of Arm and a big part of how we see the infrastructure evolving, going forward.

Geof: And you talk about edge computing there. Another big component around that, of course, is 5G. So we’ve heard a lot about 5g for years, and the impact that it will have on the world. So is it finally taking off?

Chris: Well, I you know, I think that it’s not a destination, it’s a journey, as you’ve kind of highlighted and for sure, there’s a considerable amount of 5G installations that I’ve heard happened. And you know, on your phone, you can see that your times, you’re attached to a 5G tower. But 5G has a lot of different underlying technologies and what you’re just starting to see is some of the 5G technologies that are potentially some of the more transformational technologies. And those are things like millimeter wave, which provides you relatively short distances, but can give you very high bandwidth, when you talk about you know, gigabit or above types, speeds, we have a whole evolution that’s happening in what would be called the, the kind of the mid band or kind of the sub six gigahertz band where, you know, traditionally, cellular bands were mostly sub three gigahertz. And now we’re moving into kind of the sub six gigahertz band, which puts a lot of pressure on actually the compute requirements, because generally, it’s pretty hard to move things around like cellular towers, we’ve put those have been in the, in the ground for quite some time. And so we try to use computing to overcome some of the facts that that you know, the propagation at those higher frequencies is much more difficult. And so we’re looking at using a lot of additional computing, to create the allow those signals to travel at a, you know, at a far distance with a lot of data being able to be encapsulated inside of it.

So there’s a lot of different movements around 5G and then there’s also quite a bit of discussion around some new concepts around things like private networks and kind of other business models that people are looking at enabling around 5G, so it’s kind of a barge wrapper of a whole set of technologies. But I think, yes, we’re starting to see the beginnings of it. But I would say it’s very much the early innings and really, some of the more exciting transformational stuff is yet to be deployed.

Geof: So and as you move towards that, what role do you see infrastructure playing in making it more mainstream?

Chris: You know, there’s definitely a lot of new use cases that are being discussed. Everything from you know, connected devices are connected IoT devices, ideas of how healthcare and other types of technologies could take advantage of this proliferation of connectivity. And then of course, the autonomous everything, whether that’s autonomous vehicles, autonomous robots, really trying to solve this kind of last mile problem for delivery of humans and other and other goods, as well, as there’s, you know, different types of models that are coming up relative to how that how that equipment is procured, whether that be you know, through a traditional supply chain of large telecom providers, or some of the new open models that are coming out around things like Open RAN where the idea of you know, 5G gear can be provided more without kind of a white box approach. So a lot of different angles that are that are being looked at but the one thing that’s pretty common is again this idea of software portability that I mentioned before, and so moving the software to be portable, which generally means making it being cloud native, is allowing for this software defined edge, where you can start moving around these workloads, based on the requirements either from a workload point of view, from a cost point of view, from a power point of view. So we see more and more of that being driven, as well as a big initiative of how you deploy 5G.

Geof: And that sounds like a great segue to talking about today’s announcement of the launch of the Arm Neoverse V1  and N2 platforms. Can you tell us a bit more about both of those?

Chris: Sure. So N1 was our first processor that I mentioned, that’s being used by AWS, Graviton, to and also Ampere Altra and many other kind of embedded design wins. And so we It was a great start for us and really started to open up people’s eyes and the types of performance that anyone could deliver, you know, very much in line with what was seen with from traditional architecture performance. So what we’re really excited about is with V1 and N2 is we really feel like we’re taking that to the next level. And so V1 is really about performance and it’s designed for high performance computing, for the cloud, and with a real focus towards some of these new AI and ML or artificial intelligence, machine learning assisted workloads. And so the focus is really processors being crunch all that data as quickly as possible. And we’re really excited about some of the V1 initiatives that we’ve had. It’s actually our first core to offer Scalable Vector Extensions, which we call SVE. This provides a real uplift relative to vector performance and this is also being seen quite a bit in the HPC world where they’re pairing that with very high bandwidth memory and even potentially high bandwidth HBM memory as well. We’ve got a lot of interest from the hype HPC market, everything from Sipearl in the European semiconductor Consortium, we’ve got ETRI in Korea, as well, as we recently also announced that metti in India, so all three of those country initiatives have announced support and product plans around using V1 in their domestic HPC device.

And then that kind of brings us to N2 and so N2 plays in a broad range of use cases. Again, continuing on the theme of Arm being able to provide this really this really nice sweet spot of power but performance and the amount of performance that we can pack into a very high density SoC. So for example, scaling up to 128 cores in a single device. And this whole scalability of these independent cores is something that’s been quite successful relative to how the hyperscalers are thinking about building their business and how they’re providing all these user applications with, with systems that people basically rent these cores, or you know, rent cycles on these cores is the fundamental building block for cloud computing. And so really, the fact that we’re able to offer this high density of cores that are still very, very performant in a given power envelope is quite powerful for their business and their business metrics as well as the way they can price and provide those solutions to their end customers. We’re seeing N2 being used and as a preferred core for much of the 5G build out, as I mentioned, kind of in in both networking as well as in the 5g components, such as RAN and baseband. In fact, Marvell technologies has announced that they will be using the end to part for both in data center deployments in their DPU, as well as in their next generation RAN Octeon device. So really excited about the traction that we’re getting there. And again, that’s really what Arm has been known for is this, this real, beautiful balance between high performance and it really unlocks the amount of computing that you can provide in the most dense designs, or the designs that are closest to the edge that have a suit a very severe and very difficult power envelope. And then the other thing that we are providing is we’re also providing an interconnect what we call CMN 700, which allows these high density cores to be performant. So, you know, not only do you have you know, this these, these large number of cores would be able to put them in a fabric that allows the system for them to all perform well still in that system sharing the system memory and the other system components.

Geof: Finally, how do you see infrastructure changing in the future? And what impact will it have on the rest of computing?

Chris: So I think of infrastructure as really the silent backbone across many of the technological advancements that you’re going to see, you know, today and kind of going forward, and that’s everything from drug discovery, as a kind of mentioned, to the fact that, you know, people expect connectivity and people expect, you know, whether it’s internet access, or whether it’s access to cloud computing, it’s almost become as essential in our lives as you know, electricity. And how we feel about when electricity goes away and kind of some of the inconveniences there, I think we feel very similarly about many of the infrastructure services, technological infrastructure services that that we get. So it becomes a really essential service. And yet, it’s a service that continues to grow tremendously, because of the opportunities and the new applications. And again, there are many applications that that that are really important to us and deeply shape our lives. Obviously, healthcare is one that comes up a lot in how we visit a doctor? How do we interact with experts? And how do we deal with, you know, things like our elderly? And, you know, is there better ways that we can use data, sensors, assistance, autonomy, all these types of technologies to make life better and make life more efficient, or manage costs for some in some of these areas. So, you know, I see it as this silent backbone, as I mentioned, that really allows greatness to happen on top of it. But the common denominator is that whatever amount of compute that we’re able to provide, you know, the applications will expand to, you know, utilize all that and ask for more. And so we need to be continuing to look at how do we continue to build a more performant and more efficient computing infrastructure. And that is what Arm is very focused on. And, and we think we’re doing quite well there. But we’re really excited about what the future holds, because we think many of the transformational technologies like ML and AI are just starting to kind of scratch the surface of what we’re going to see going forward.

Geof: Thanks so much for that, Chris. And I’m sure we’re going to be hearing a lot more about where these innovations in infrastructure are going to take us. We look forward to bringing you more news in the next episode of Arm Viewpoints. And look forward to connecting with you all again soon. Thanks for listening today.