Transcript
the agenda because as you guys heard in my opening remarks, we like to think about the future here at scale. And so we brought these guys together to discuss that. So to my right, you saw him earlier, it's Brian Chambers, Chief Architect from Chick-fil-A and also author of the Tech Chamber of Secrets. Steve Henshaw, Senior R&D Engineer from In-N-Out Burger, and I will point out as a huge fan of fast food, I strategically positioned these guys nearest me. So Flavio Bonomi from Accenture. We've got Aaron Roswell, the founder and president of Simply Nook and our very own Dave Demlow, VP of Strategy at Scale. All right. So let's get started. And I thought where we would start, when we saw Brian's keynote earlier, one of the things that caught my eye was the Apollo computer and the miniaturization and thinking about, you know, in those days, like you pointed out, I mean, those, you know, even pre-mainframes, early mainframes were entire floors of buildings kind of thing down into something small. So, you know, just from a, even starting from a hardware perspective, and it's probably the best question for Aaron, right? You guys have been innovating around miniaturization for a long time, right? We saw the, you know, kind of your original product that had a little bit of a sci-fi look with that chrome finish. But, you know, tell us a little bit about how you're thinking of the future of that miniaturization. And also, more acutely, for those in the room, you know, how you guys have helped position those, that tiny PC form factor into something more than just a hobbyist kind of device. Right. Yeah, so when we came out with the space-aged mini-computer, it kind of put me in a unique situation because we couldn't produce it at a cost that was, you know, feasible for hobbyists and consumers. So what we started to see is a lot of adoption from commercial applications and even, you know, Fortune 50 companies reaching out to us saying, hey, I need a small device that's powerful that's going to fit in this specific application. So we kind of got an early glimpse into where we could go if we could get a powerful compute device down into a very small footprint. And the applications almost were endless as to what you could do with the product. So when the Intel Nook came out, I kind of saw this, you know, these guys are doing all the hard stuff, which is mass producing them at scale. And so, you know, using that to start Simply Nook propelled us into this world of our customers showing us what they could do with these small devices and how powerful, small footprint, low power devices can be used in so many different applications. Yeah, it's been interesting to see. Where do you see kind of that trend going? Right? I mean, so probably not just smaller and smaller, although that doesn't hurt, right? Right. Yeah. So, I mean, we're showing off a few devices here that we haven't even announced yet that Similarly to what we did with the PC and shrinking it down. We're trying to do with servers now. So same type of functionality, same type of enterprise level features and capabilities in a small footprint, but extremely powerful and kind of more of that AI ready device that's 24 seven operation and can be trusted in an enterprise environment. Yeah, very cool. I mean, one of the things that, you know, every partner in the room and every IT professional in the room is challenged with is trying to predict the future, right? And those that are successful at it propel their organizations forward and those that don't tend to get left behind. And one of the things I think, you know, that I thought of while you were answering that and I noted earlier is that even in the early days of the Nook where, you know, you weren't using terms like, you know, a miniature enterprise server, you know, but both of these guys here, you guys were using those, you saw something, right? You saw something in those Nooks that said, oh, I could use these in my environment. So maybe kind of just reflect on what you saw then and maybe what you think in the future. Yeah, I mean, I can start. I think there's two things, you know, history kind of repeats itself. And I think you can learn a lot from looking at the past. And like completely honestly, a lot of our inspiration for picking a small form factor and something consumer grade as opposed to something enterprise, I mean, there's a lot of factors, cost, all these kinds of things. But if you remember the early days of Google, Google actually started with a lot of consumer grade hardware that was chained together to enable a lot of their early, you know, search. Legos, they had Legos, they used the standoffs. Yeah, yeah, really, really interesting. So they took something very available, very low grade, but they were successful by using lots of them to accomplish the larger task. And if I ask myself, given a budget, would I rather have one thing that's really, really good and powerful, or three things that are a little less so, but more redundant? I'll take the redundant, you know, every single time. And so that was a thought process that brought us to that. Not only that, we also have a constrained physical space in restaurants. We don't have a data center closet or anything like that, even. We're not dealing with racks, you know, air conditioned spaces. So we needed something that would fit our space and that would work in that, you know, that office type environment. So it was a good fit for us in that regard. Yeah, that's kind of like what brought me to that line, too, is because back in 2014, I saw that line of computers and it's like I needed these little workstations. And instead of getting a whole tower computer, it was like, OK, well, I can just put this in there. It's going to do it. It's got enough horsepower to the workstation I needed it for. And I just keep following that line. And I found it using digital signage and I can use it over here. And then now it's on the cookout truck. So it's like it just kept evolving and I just keep going down that line of hardware and I love it. So, yeah, I mean, to borrow, paraphrase, you know, Brian, one of your comments, I think, you know, make it as small as you can, but no smaller. It's that kind of a thing. So, so, Aaron, you were talking a little bit about, you know, the future of the devices and A.I. and obviously A.I. is a big, big driver of a lot of what we see around edge. You know, Flavio, we were talking earlier today about some of the industries where we're seeing A.I. and obviously at Accenture, you guys get visibility into all sorts of industries. So so maybe just kind of opine on, you know, where you see A.I. taking hold now and where will it go in the near term? Well, A.I. is is the hot topic these days, particularly after open A.I. came out with the generative A.I. opening up and the the beauty of this is that it is the forcing factor towards a continuum of computing, of data distribution and of applications, distributed applications. It is really requiring that we we train in the in the cloud. Many times we we we updated and infer at the edge and there's this need for a more dynamic use of all these resources so that not only we have much more powerful, small, miniaturized computers, but we have a continuum of computing that we can now use. And there is where also scale excels in terms of organizing and orchestrating these resources. So what I see is really the bringing of this A.I. at the edge. Edge A.I. is a big forcing factor for edge, I believe, at this moment because it's the hard application that can really help not only on visual visual applications, but the generative A.I. local using local data is the big is the big element that it can help people access documents, respond to questions, answer to problems, operating operational problems and others. So and in fact, Intel is pushing this A.I. everywhere. And I think we will see in the NUCs, in the PCs, in the phones is happening and will happen and all the way through. But not only A.I., I think with the A.I. comes digital twins, which are very similar, but but complementary that, again, require this distributed infrastructure and can help generate data, understand data more, sometimes more reliably than A.I. But the movement is clear. And you touched on something there, which I think is probably useful for the audience. I think that we generically have begun to associate A.I. with GPU, but but in reality, a lot's changing there. Right. I mean, you mentioned Intel and I know that's an area that you know about. So we see, for example, CPUs capable of running significantly large models, even MCUs in sensors capable of running voice recognition, video recognition, vibration recognition, A.I. models. So I think that towards the edge, we need to to scale. We need cost, power, sensitivity. And so we have to use optimally all the ways of computing. And it's not only GPUs. In fact, it's ASICs, CPUs, FPGAs and so on. So so taking another look at I'm using Brian's keynote as kind of my anchor for I haven't I haven't asked a single question that these guys. So thinking about, you know, when you when you got started in your project, you know, you mentioned that in the back of of your restaurants and pretty much all the restaurants, there were a couple of Windows servers running things. But specifically when you set out on your edge journey, you viewed it as a green field opportunity. And obviously you talked about how you used containers for that. So I'd be interested, you know, if you can maybe reflect a little bit on, you know, why containers for that particular use case, but also if you would touch on your use of open source technology. Yeah, absolutely. I mean, the simple way to put it is I talked about in the keynote the proximity to the cloud native paradigm and where a lot of the cloud native computing stuff has gone is towards containerization. It's the way that we deploy workloads when we build them in the cloud, deploy them to AWS or wherever else. It's generally our default. So when it's possible, it seems prudent to use the exact same approach for your developers to get a consistent application development experience across all of the different platforms that you could potentially deploy to. So that was the real motivator for containers. I don't think containers are better than VMs or better than Wasm or better than anything else. It's just what is your organization most comfortable with, most skilled and most tooled around to operate? And then, you know, what makes sense to do as a result? I think you had a second part of your question. Yeah, we'll hold that for a second. I have a follow up for that. But I was thinking and maybe this is back over to Flavio, but where you took a kind of a greenfield approach, many, many customers can't, right? Many customers have a mix of applications, legacy VMs, even bare metal. So what kind of things in terms of management orchestration are you seeing customers struggle with, Flavio? And this is not a scale computing commercial, of course, where we already have believers in the room. But, you know, what kind of challenges do you see? So I wanted to link to these old containers in one shot. It is not going to happen because, again, there's a lot of brownfield there and there's a lot of applications that are not modernized and cannot be thrown away in one bucket. So there is there are two big reasons to go into an infrastructure that virtualizes at the level of the VMs first, hosting then the containers in VMs. The first one is this slow process of modernization of the applications and the need to start consolidating old and new. So by bringing in a more complete virtualization approach, you can do the new while protecting and modernizing the old that sometimes cannot be containerized all the way. The other one is when you go deeper in this movement towards the far edge, towards the physical devices, towards the cyber physical systems, it's a world of real time of control and virtualization with containers is not there yet for that. You would like to have a mixed criticality environment separated where applications and domains are separated and secured through virtual machines, through a good hypervisor. So I think that two big reasons I am for this kind of progressive movement and I appreciate now this is my pitch for scale, scale focused exactly on that area. That is the first area that needs to be brought in. That is the foundation, the platform, and on that you can really play. I think one of the cool things about what we're doing with infrastructure as code and so forth is that you can take those legacy applications and still put them on more of a modern path in terms of how you're managing them. So Dave, maybe you can think a little bit about that. Again, these aren't prepped questions, so I'm talking so you have a few seconds to think about a response. But I mean, you've talked about GitOps and Ansible and so forth, and you've seen customers using this for both containers and for legacy VMs. Right, right. And that's definitely something that we feel is important is to provide as much of that cloud native approach, cloud like experience, API driven infrastructure as code, regardless of what's being deployed. You know, another kind of interesting thing on the pure containers everywhere is we see a lot of customers where it's not just one development team. It's not one kind of pipeline. It's hybrid cloud, you know, multi cloud. You know, some development team picked Google and is running there and has certain processes. And within the same company, another organization or an outside vendor is developing an app. So, you know, we're in the position where we wanted to make sure that we could promote those opinionated best practices where applicable. We could work with whatever's out there, you know, so definitely brownfield environment from legacy VMs to different types of container runtimes, different container orchestrators, different cloud control planes. You know, where do you want to do you want to drive things, you know, from kind of an edge in or cloud out? We want to be able to support and integrate to either of those. Yeah, I mean, you brought up a good point that, you know, and this is a bit of a scale commercial maybe, but, you know, we've taken a bit of a Switzerland approach to how you're going to handle those containers. And it's very it's more common than you might even think. And this is maybe a tip for the partners out there. But you'll get a customer where the IT team or the DevOps team may have selected a particular cloud provider. But then the other team next thing you know is developing on another one. And so you actually need to run, you know, Azure and Google and VMs all on the same kind of platform. So that's, you know, if you wonder, you know, why is scale taking that approach? It's because we see it in the real world where those sorts of things are happening. You know, thinking about some of those tools, I mean, a lot of the integrations and the tools we're using there are open source type tools. And so that brings me back to Brian. That was the other part of my question. Talk about, you know, how you guys leveraged open source and why. Yeah, I mean, one of the reasons that we chose Kubernetes as a layer for our solution was all of the momentum that existed around it with open source tools. I mean, if you look at the CNCF, Cloud Native Computing Foundation landscape, there's like projects everywhere that do all kinds of things. And, you know, the community around things is really important to anything that you're going to build on top of. Like if you want longevity, you know, if you want things that are going to be supported, that could be commercial or open source. It doesn't really matter. But you want a thriving ecosystem, right? And that's what we saw around Kubernetes. And we've taken advantage of a whole bunch of open source projects within that ecosystem specifically because it's the one that we work within. But it's a big part of our Edge solution stack from our MQTT broker to, you know, we built on a bunch of open standards as well, which are open source things like OAuth from an authentication authorization perspective and on and on. So I think it's like it's a foundational component that you can choose to latch onto and build on top of. And, you know, it's a big force multiplier when you do that because it's concerns that you don't have to go solve for yourself. So thinking about that again, totally unprepped questions here. But Dave, obviously, you know, we leverage a lot of open source at scale. So as the VP of strategy, when you're thinking about, OK, what tools are we going to use next? I mean, what are how are you thinking about that? Yeah. Community is the big part. And, you know, I'm Kubernetes. We've had a lot of internal discussions about, you know, what kind of container oriented runtimes do we support? Do we support all of them? Do we, you know, get more opinionated? And I do think there's an opportunity for both. But there's just so much momentum around Kubernetes. It does. I mean, if you really look at, you know, what's at the heart of Kubernetes, it's a lot of the things that we've built into our hyper core operating system, you know, state machines and conditional processing and, you know, AI ops essentially built in, you know, that are, you know, driving towards desired state operation. So there's a lot of in that case, if there's good stuff out there to leverage, even contribute to. I think that's where we as an organization want to you want to reinvent the wheel. We don't want to create friction when there's, you know, good, good options or even, you know, multiple competing options. That's one of the great things about Kubernetes is, you know, and most standards, if you don't like the standard you have now, pick the next one. And there's lots of ways to, you know, to do things. Yeah. I mean, and things that we've done at scale that would be completely invisible to the audience. You know, we have engaged in open source projects because we thought we were going to use them, contributed to those projects and then never actually did anything with them. Right. And that's just part of it. Right. You kind of move on from there. So, you know, now both the guys here to my right are in the retail industry. So I thought, you know, maybe, Aaron, I mean, you guys obviously sell horizontally into all, you know, anywhere from I want to build a set top box to use it commercial. But, you know, what are maybe some of the industries where you see your devices being used more than others? Yeah. I mean, it's almost endless. I'm always surprised by, oh, wow, this thing's being used for, you know, gosh, now I'm drawing a blank. Guys, help me out here. AI is a big one. That's an up and comer, you know, AI drive through. We've seen that coming up where they're doing predictive stuff in drive throughs. We're seeing, man, I'm. Well, I saw a thing of a drive through. I saw this interesting transportation application. It was it was for restaurants again. So it's kind of about these guys, but not quite, because this was restaurants that sold beer. So but the app is they were using computer vision to optimize the moment in time at which the waitstaff should ask if someone wanted another beer. So the waitstaff wore smart watches and based on where the beer was. And the claim was that they could raise sales, you know, 15 or 20 percent. We'll see. Right. If that's if that's true. So, you know, one of the we see at scale, you know, really two big markets right now embracing both edge and and a number one is retail, which is hence you guys are on here. But the other is manufacturing. And I know that that falls into your wheelhouse, Flavio. So what kind of things you've seen in manufacturing? And I know we would likely have many partners in the audience that sell into large and small manufacturing and the PLC areas of particular interest. Well, manufacturing is a is a big play for this modernization of the of the applications and of the infrastructure. And I think we are seeing the understanding that the future is coming in from this cloud native approaches and and the openness is there. But again, what opened the doors is this future with AI, AI for inspection, AI for protection of workers, generative AI for better responsiveness and so on. So that this need for to bring in the edge is opening the door. And now we see the beautiful coexistence of an openness to this coexistence between the old and the new. What is really impressive for me is to see that there's openness to go even deeper, you know, to go into the embedded all the way through. And that touches the real control. And I feel that control has been stagnant in many ways for many years from the invention of the PLC. We have done a lot with it. We did a little bit of distributed control, but it's not really it has not made big progress in many, many years. Now we have the ability, the potential to bring intelligence to the control, to bring these digital twins models with high performance computing to decide better, to choose the better paths for controls, to organize the control, aggregate the control and optimize, for example, energy consumption or process flow, process efficiency. So there's a huge potential. The intelligence is coming in with over the new infrastructure. The openness to virtualize the controllers is coming in. The next step will be to embed this intelligence in the loop, in the control loop, which requires the network attention to the network and advances in distributed control, which is not fully there. But but this is the big goal and the big nirvana that will improve efficiency dramatically and bring autonomy in many of those decisions. Yeah, I think it's so interesting because I know we have, you know, we have customers in the room who in the manufacturing sector have PLCs, the PLCs produce data. They're running NUC clusters from Aaron, very close to those PLCs, not on the normal industrial network, but on their own kind of mini network to keep the data off. But but then that gives them a platform to run other applications right there near the data. And certainly thinking of futures, you could see where, you know, especially as you've got, you know, AI, special AI chips and being able to do things with CPU running those those workloads right right on there. You know, Aaron, I was thinking about that use case and these guys. And so, you know, when you guys deployed in your restaurant environments and used sort of off the shelf NUCs, right, and restaurants aren't data centers. So, you know, there's probably a little more grease in the air and temperatures a little different, you know, on the industrial side, you know, not typically pristine data center like environment. So, you know, what what have you seen, Aaron? And I know you guys have ruggedized type systems, but, you know, you obviously sometimes see customers running, you know, regular NUCs in these environments, but kind of talk about that that mix. Yeah, we like to understand the environment just so we can recommend, you know, not putting something with a fan right above the deep fryer. It's not going to last as long as you would like it to. But, yeah, it really, really depends on the environment where they're at. Luckily, Chick-fil-A has them, you know, kind of back away in an office that's nice and protected. But you do get into these scenarios where it's it's it's maybe it's outside, you know, or in a auto mechanics bay and the doors open and it's 100 degrees outside and there's dust and debris flying around. So we try to tailor our devices specific for certain applications where they're going to work best in that environment back to these kind of extreme edge servers that were where things have kind of led to going forward. Almost all of those are designed with kind of a fanless enclosure and no no active cooling. And again, recognizing that, you know, even the best laid plans go awry. Right. Even when you put those ruggedized systems in, sometimes the grease still gets in there. I mean, Dave, you I know you use the phrase cattle versus pets. Right. So, yeah. Oh, I got four minutes, it says. Oh, I have a meeting to do. All right. Well, we got it. So we're going to I'm the one messing this thing up. Go figure. All right. So I better get to the last question, which is the only one I've asked the last that has been written down. So why don't we, you know, very quickly go down the line, just give us, you know, 30 seconds or 60 seconds on if we were on this stage, you know, five years or 10 years from now looking back and what do you think we would see? I think more edge. If you remember the Gartner slide from yesterday, it said that about 50 percent of businesses either don't think they need it or don't have any plans to use it. And I think that'll change. And I hope as it happens, there's more standardization, more best practices, more consistency in the industry about how to do it, how to do it well. For me, it's like I'm a hardware guy, so I really think with all this stuff coming at us and stuff like that and the compute just is going to get more and more powerful as it always has. There's got to be a better way to keep these things cool because that's the biggest thing I see on the hardware side is how do I keep this thing cool? So advances there. I don't know what it's going to be, but something. Put it in your ice machines. For me, given the speed of the progress, I've been at this topic for many, many years, so I don't want to say this is closed. So my future view, three, four years is what I just said. More edge at the real time boundary at the boundary with the physical systems embedding more intelligence so that it will take time. We'll be there. I think we're all on the same page there. I think Michael Dell has a quote and I don't know the exact quote, but it's how big the edge is actually going to be. And none of us really realize how big it's going to be and with how fast is accelerating things. I posted a thing today on my LinkedIn. Just I mean, in the most recent few months, we've made huge strides on on mathematical AI calculations. Right. So a few months ago, they weren't so great. Now they're almost better than expert mathematicians can do. So, yeah, I just think we'll learn more and more about what we can do to be more efficient out at the edge. And that'll require more compute. And, you know, that's a good thing, I think, for all of us. Pretty much all of that. I mean, you know, to maybe add some other things, you know, we started talking about miniaturization. It's not just been the compute, but it's the sensors, the devices, the ability to collect tons of data. And I think there's still room there, like there's still untapped data. I also think, you know, in AI, you know, there's lots of the obvious, you know, computer vision applications. But, you know, I'm starting to hear and see cases where, you know, different different point solutions are being combined together and leveraging each other. And, you know, so it's not just the vision detection, that object, it's, you know, we also know the customer behavior and putting all those things together in kind of a way that's, I think, just starting now. And so, yeah. All right. So to summarize, we're going to collect a lot of data at the edge. We're going to run on embedded systems and we're hoping not to burn the place down. All right. How about a round of applause for these guys? They did an awesome job. Thank you, gentlemen. I'll see you guys later this evening.
