25. Hardware Still Matters at the Edge – Tech Field Day Podcast

Welcome to the Tech Field Day podcast, where we bring together it industry experts to discuss topics that are important within the industry today. Often we do this in association with our tech field day in person events. And always we distribute this content all through Tech Field Day, as well as through our wider future and group, and particularly our friends at Techstrong Media.

We often record these podcasts in association with the in-person events. This podcast is associated with tomorrow's Edge Field Day event that's happening here in Silicon Valley. One of the things that I've seen in the collection of presenters at this Edge Field Day event, as well as some of the previous events, is that there's a huge amount of innovation going along in server hardware or infrastructure hardware at the edge, some innovation that I haven't seen in server hardware for quite some time in the data center.

And it's driven by those unique requirements for running applications in these diverse edge locations and these increasingly diverse edge locations. Joining me for this episode of the Tech Field Day podcast, I have, of course, uh, Jack Pollard. Jack, uh, welcome to the podcast.

Thank you, Alistair. It's a pleasure to be here. Whereabouts Are you from my friend?

Uh, I am based on the East coast of the United States, and I am founder and principal analyst of Paradigm Technica, an industry analyst firm. Excellent. And of course, no stranger to the Tech Field Day podcast as my other guest.

Mr. Stephen Foskett. How are you doing today, Stephen?

Well, I'm actually really enjoying this because I get to be on the, around the table as a, uh, commentator instead of the host. This time on the Tech Field Day podcast, uh, Alistair and I are gonna be leading the Edge Field Day event this week together in anticipation of Alistair, uh, leading more field day events going forward. And of course, all three of us will be attending Edge Field Day and, uh, covering all of the presentations this week.

So I did want to start the discussion with the innovation that we are seeing from these hardware devices. We have three different vendors who are fundamentally about the hardware devices at the edge, and Jack, you and I were discussing some of the reasons why we're seeing the innovation. Yeah.

One of the big things is when you look at enterprise, it, everything's driven by commodity. There's such volume of systems out there, but they're all cookie cutter. They're all effectively the same.

So prices keep going down and we build more and more and more of the same boxes. When you look at the edge, what's driving innovation is the need to put systems in various different locations that have different physical requirements, different operating requirements, and the physical plant, the network, the power, the cooling, everything is different and it's not standardized like it is in a server, in a, in an IT data center. Right.

And that's, that's actually one of the things that I love the most about the edge topic. It's one of my sort of, uh, hobby horse kind of pod, uh, topic events that, that I love covering because it is so fun and different and unique. And as you mentioned, um, these unique hardware aspects are driven by the, the special nature of the edge because essentially, if, if you're trying to define the edge, the edge is everything that the data center in the cloud is not.

I, I, you know, it's, it's easiest to define it as everywhere else. Um, and, and that everywhere else is a really fascinating aspect because it means that you can't just use the same stuff that you use in the data center because everywhere else could be, you know, as people, like in the, in the industry, like to say, it could be the top of a windmill, it could be a main battle tank, uh, deployed on a battlefield. It could be a grocery store, it could be underneath the fryers at a quick serve restaurant.

It could be almost anywhere. And so it is very cool to see how, uh, hardware, which essentially, I mean, it's all PCs, of course, it's all typical PC hardware conceptually, but it is radically different in reality. Mm-hmm.

I think one of the key differences I see is that when we build servers to go into a data center, we build the environment to be friendly to those servers. We have air conditioning, keeping things at a regular temperature. We have reliable power.

We even try and avoid vibrations. And that's not something that you can do in these edge locations. Uh, as Steven says, the categorized as being not a data center, and so it can be all kinds of hostile environments.

The vibrations in that main battle tank, uh, are going to be significantly more severe than most of the hardware we put into data centers would cope with. And so the design paradigm shifts from being we build an environment that is going to be the most conducive to the long life, and the ease of deploying these relatively large pieces of tin that we put into data centers, we say it's gonna be hostile environment, and your hardware has to survive that hostile environment. It's not just physical hostile.

It's one of the things that struck me as we were talking to some of the vendors at previous Edgefield Day locations. It can also be hostile actors, human beings who want to steal stuff or tam with this stuff. These aren't designed decisions we have to think about in the data center.

No, they're not. But they're also not new issues. This is the other thing that I find fascinating is that, you know, 30, 40, 50 years on in the computer industry, we're still doing the edge.

We think about it as new, and it's a new term, but I remember when I started out in the industry with many computers putting them into grocery stores, and it was a tiny computer in a very large box. It was the size of like, what we think about a dorm fridge. And people would ask the manufacturer, why is it so big?

Right? It's a tiny little computer inside this big box. And the reason is 'cause it would get stuffed into the manager's office and he would use it as a desk, right?

And so you had to have enough cooling and enough protection around it. So when they spilled the Coke or whatever, or you know, they sat on the thing, it still functioned. And these considerations have been around for a while, but we're now coming back to it as things are driving an explosion in the need for edge.

You know, and we're thinking about, you know, what's the, the, the big elephant in the room is ai, right? And how are we gonna start moving AI out to the edge into these remote places or hostile environments or in the military or, you know, oil and gas exploration where you're out in the middle of nowhere, uh, doing a pounding the ground and doing seismic studies, and you'd like to be able to process that data immediately so that you can figure out where next to go, rather than having to bring petabytes and petabytes of data, ship that all the way back to a central location to process it. And that's very time wasting and say inefficient.

So we're pushing more and more stuff to the edge for efficiency and then having all these other requirements, right? Yeah, and I, I, I find it interesting too to see how much of what happens at the edge is inspired by and sort of riffing on what happens in other areas. Mm-hmm.

So for example, uh, the physical form factor of edge servers very much reminds me of the physical form factor of, um, industrial computers. Sure. But specifically of wifi and, um, wireless, uh, you know, generally, uh, fixed wireless devices, um, that have, uh, you know, they have big heat sinks for cooling and die cast chassis for durability.

And they try to try to make do without fans. Um, they use power over ethernet, things like that, you know, DC power, things that you just don't see in data center environments. And the same is true of so many other aspects of the edge.

Um, you know, when it comes to, uh, AI capability, a lot of that is actually driven by mobile devices, uh, laptops and phones and so on. You know, the, the fact that they're, the, the chips are getting more and more, uh, neural processing capabilities that we have low powered GPUs that are designed for laptops, but can be used in edge devices in many case edge servers are actually very strange little laptops because, you know, they are using the same technology when it comes to iot devices. A lot of that is driven by, um, mobile phones, um, technology.

So since phones have things like accelerometers and incredibly good cameras and, you know, all these ultrasonic sensors and touch sensors and all these, all of that stuff can be adapted for use in edge IOT environments. And when it comes to software as well, uh, virtualization a a absolutely has transformed the edge from a very, uh, sort of one device per use case to multi-device kind of situation. And now we're seeing the same thing happening with Kubernetes, but Kubernetes, just like the rest of these technologies, is not being deployed the same way at the edge that it would be in the data center.

It's being used entirely different. It's like a tool they found in the toolbox and said, wait, we can do something cool with that. And so there's this whole kind of collision of things that come together and get kind of remixed in a different way.

And that's just fun. And I think you bring us back into a, an important point that is not just about the hardware, hardware, it's about a synergy between the application that's required to be delivered. And in the end, we're all about delivering an application out to the end point where this, this, uh, edge technology is.

But there's that synergy between the hardware, software, and the actual business application. And we do see that Edge locations tend to be getting a little more general purpose. So they started as single purpose.

You have a piece of hardware that is specifically for one application and does only that one application, but we're starting to see more of a dynamic move that there might be new applications that are being delivered in the lifetime of this hardware. And so we need more of a general purpose platform for running these applications, often containerized. And often now we're bringing more AI applications in.

But there's some challenges with this. These aren't the same sorts of servers that we're deploying this software into that would have in our data centers. We have these things in intermittently connected locations.

And distribution and control can be a real challenge. Well, I think what, what we as in an industry have to think about is just as you do for hardware on the software side, you have to throw out all of your assumptions. When you're on the hardware.

You assume that you have great network connectivity, that you have great power, you have great cooling, and all of that is taken care of for you. And as a software person, you can just go and write With great power comes Great responsibility. But there we go.

Um, in on the software side, it's the same thing. It is assumed that you have a set of tools available to you that can do what you need to do. And you, and you just sort of go bli marching along, building your application with knowing that you have Kubernetes and that it works the way you think it works in a server environment with multiple processors and large memory systems.

And when you get out to the edge and you have limited computational capability, you have limited network comp availability. You may not have ECC corrected memory, right? So you can have memory errors, you might have intermittent, um, uh, even box to box connectivity in the same, the same edge data center, so to speak.

You might have connectivity issues for physical reasons, right? So all of these assumptions you have to throw out the window and say, let's start from scratch and understand exactly what our environment for this application looks like. And then what tools, as you said, like Kubernetes, what tools are available that might help us build what we wanna build in that environment.

And that drives innovation. And that's why, you know, I'm so looking forward to hearing from these vendors because it's gonna be so interesting all the different places that they're innovating, where in the enterprise data center, we're like, there's nothing new here. There hasn't been anything new for a while.

'cause it's all cookie cutter To some extent. But as, as you and I saw too, you know, there are still companies that are doing cool hardware in the, in the enterprise. There are absolutely, Uh, you know, we saw Oxide computer, which was absolutely incredible, but that's very rare.

It is. And for the most part, even special purpose devices like network switches and security devices and storage devices are using commodity hardware. And it's true that in the edge they use commodity hardware too, but they're just using it in so much more interesting ways.

You know, it's, it's incredible to see how they're building, um, storage systems that are reliable, high performance, high capacity, you know, flexible, you know, they, they're able to be sealed against the bad, you know, environment right? Against, uh, vibration and shock and so on. And, and, and because data collection is so much more important with this advent of ai.

And so I kind of wanna get to that too. Like AI at the edge is probably the big topic that's emerging right now. And everybody, just like in the data center and the cloud and personal devices and everything, everybody's trying to talk about how they're gonna do AI at the edge.

And from a software perspective, I think we can assume that we're all still kind of figuring it out, but that it's gonna kind of look a little bit like what we would do on mobile devices and in the cloud and so on. But from a hardware perspective, it's, it's totally insane because we don't have great power. Yep.

You know, we have to use low power, we have to use special processors, we have to optimize the data paths. And I think that's another big aspect that's, that's gonna kind of collide on all this is that as, and this was one of the things we talked about on the utilizing AI podcast or utilizing tech early seasons, as we have more AI to process data, it drives us to collect more data. Yes.

And that just means that we have more and more and more and more data that we have to work with, which means that we have to have more capacity and also more, um, throughput and, uh, for that data, you know, more ability to move data around. That's not something you can take for granted anywhere, but certainly not when you're talking about small compact, low powered devices. Well, and you know, you talked about, you know, we're still using commodity.

And I look at it and think about it this way, we're using commodity architecture, right? Mm-hmm. A computer for the Edge still looks architecturally like a computer in a server, in a data center.

It's still got a processor, it's still got memory, it's still got network connectivity, it's got some storage, but we're gonna take that architecture and maybe have that architecture as a laptop and repurpose that in a way that the manufacturer or, or the vendor never really thought of to begin with. Like, let's lip the rip, rip the board outta this laptop and put it in a case without a screen, without a keyboard. But everything else stays the same.

But then we can protect it and use it in such a way that we can do AI computation on it, where they, you know, this was the, the laptop was originally designed to do Word, right? Mm-hmm. And email, and that's in web web browsing.

But now we're gonna do all these other unnatural acts on it, and so we have to innovate to figure out how to do this, right? Yeah. And I think that innovation we're already seeing in the laptops as, as a, as a thing.

The A IPC is a real thing, and we're seeing hardware offloads inside laptops, and that technology is going to bleed into the edge to allow us to have that hardware acceleration of, of AI applications at the edge. Yeah. And, and that absolutely is coming.

And I think that that's something that's gonna be a huge topic for companies that are making edge hardware. It, you know, how do we fit AI into the envelope that we've created? And there's, you know, many elements of the system that are gonna change because of that.

Number one, power consumption. Mm-hmm. Uh, you're gonna have to supply it with more power, which means that you may have to change the fundamental power supply that you use for these devices.

Uh, number two, uh, cooling, you're gonna have to have much bigger and more advanced heat sinks. And perhaps you're gonna have to add a fan where you didn't have one before. Uh, you also have to think of, uh, bandwidth inside the device in terms of moving data around, uh, storage capacity is gonna, is gonna take up more physical space.

And all of these are going to challenge, let's say the form factors. But yet at the same time, the cool thing is this is the IT industry. And in tech, things just keep getting smaller and lower powered and, and, and more capable f And so, you know, these, these, these things are kind of coming together.

I think we're gonna see some incredibly powerful devices coming down the road very, very soon By the, by the same token on the software side again. So there's assumptions that have to be broken, and we have to rethink a little bit of how we do things, right? So you talked about virtualization and virtualization is gonna help a lot, but a lot of what we do with containers and virtualization and application development assumes we have great connectivity.

And we think about things like, uh, if I'm waiting for data, I'm waiting for data for a millisecond, not I'm waiting data for waiting for data for 10 seconds or 20 seconds, or a week or a week while it, while it makes my grades its way from where it is to where it needs to be. And the, you know, there, there are sort of built in assumptions of performance of storage systems or of network systems that when those assumptions break, the entire application breaks and becomes very hard to debug and figure out why that application isn't working the way we assume it is. So, you know, the, the infrastructure around it, the tooling, the, the, the how do we figure out what's broken and how why it broke and fix it, all of that is also gonna end up having to be innovative.

We're gonna innovate around that and provide new ways of figuring these things out as well. And so this is sort of like, you know, we think about AI as sort of the next frontier of, you know, on the, the, the enterprise IT data center side. And I think the edge, just everything about the edge is gonna be another frontier where things are gonna change and keep innovating and do things differently.

Well, thank you all for joining us on this episode of the Tech Field Day podcast. Before we close things out, where can we catch up with you and where can we continue these conversations? Mr.

Foskett, where can we find you online? Well, you'll find me as s Foskett on pretty much everywhere, uh, social media networks. Uh, I'm pretty active on Mastodon, along with, uh, uh, increasingly, um, looking at threads thanks to, uh, my good friend Karen, who's convinced me that that's a good platform.

Uh, as well as of course, uh, the Textron sites. You'll see me on Textron Gang most Tuesdays. Um, you'll see our, uh, content appearing, uh, across the various Textron sites and Gestalt it And Jack, where can we find you?

Uh, well, you can find me on my social media as Poller and, uh, but I'm most active on LinkedIn as Jack Poller. com. Of course, you can find me Alister Cook as at Dimi Tess in Z, and on all of the usual social media around it.

A little more active on, uh, LinkedIn than I've been recently. Of course, tomorrow we're starting the awesome Edge Field Day, and you really should keep an eye on Tech, tech strong media and the, uh, LinkedIn for Tech Field Day, as well as all of the other usual places that you hear from us. Also watch Twitter for x for a little bit of coverage from the delegates in real time.

It can be interesting to see their thoughts there. Thank you for listening to this episode of the Tech Field Day podcast. If you enjoyed this episode, consider giving a subscription and whichever favorite form of podcast, uh, tool you like.

Of course the listing on iTunes is easy to find, but if like me, you use Android, uh, quick Search in your favorite podcast, you will find us. You can of course also find all of this media and all of this content on the Techstrong TV side as because of course, tech Field Day is a part of the fu and group, and we're absolutely delighted to have so many new places to engage with. All of all of our friends.

com/podcast to find all of the episodes. The back catalog goes for a long way, as Steven says, he's delighted not to be the person hosting this one because he has hosted so many awesome podcasts, including with both Jack and myself as guests. Thanks for listening and, uh, we will see you next week in our next episode.