31. Edge Computing is a Melting Pot of Technology – Tech Field Day Podcast

The Edge is an exciting place full of new technologies or existing technologies that are being brought together in new and exciting ways. This really is a melting pot of technology at the edge. This episode of the Tech Field Day podcast features John Osmond, guy Courier, Ivan McPhee, and myself, Alistair Cook, uh, describing the melting pot that is Edge Computing.

Welcome to the Tech Field Day podcast, where we bring together a group of it technical experts to discuss a single topic, uh, about key concepts in the industry. This podcast features a variety of, uh, perspectives from members of the Tech Field Day Delicate, uh, community, and is often recorded in association with our Tech Field Day events. Tech Field Day is a part of the Futurum Group, and this podcast is also published with our sister company Techstrong tv.

In this episode of the podcast recorded live at Edge Field Day, we'll be discussing the premise. The Edge is the new melting pot of technology. Before the discussion, let's meet who's on the panel today, and Ivan McPhee.

I'm delighted to have you on this panel at your very First Tech Field Day event. Tell us just a little bit about yourself. Well, thanks a lot for having me.

Really appreciate it. Great to meet everybody and to be able to be part of this day. I'm Ivan McPhee.

I'm an analyst with Giga Ohm. I'm focused on the TE security and networking sectors. I'm John.

My name's John Osmond. I'm from Buerre, New Mexico. I'm a consultant.

I primarily work with networking and interconnection and things of that nature in the hinterlands. I'm guy, Uh, guy Courier. com.

I think this idea that the Edge is a melting pot of technologies is something that we've seen as we've been experiencing the presentations here at each field day. Now we're at the end of day day one. We've had three of the presentations.

I'm sure we'll see more of the melting pot next week. But we've already seen the melding of having containers for virtual for, uh, your applications along with using those containers to deploy virtual machines out at the edge. And so we're seeing a variety of technologies there.

We saw things like machine learning being actually operated on very small edge devices. Now, I really like that the example given was a conventional AI where it was image recognition using open CV inside of frigate, some of the tools that you might have been using for some time to build actual AI applications long before this large language model came along and vastly increase our resource needs. But then we saw a vast amount of resource being delivered out into an hedge.

We saw the ability to add huge amounts of storage and huge amounts of offload for AI to your existing edge hosts. And so we saw quite a lot of different technologies being glued together to answer a variety of, of different requirements. Um, John, you are particularly focused on networking.

Do you see that sort of networking stack getting pushed into the edge components and being glued together? I do at times. Um, it's a little difficult sometimes in some of the smaller cities, in rural areas that don't have a lot of connectivity.

Um, and a number of the places are starting to get, um, a lot of grant money and things of that nature to build out middle mile pieces and some last mile into their, into the cities. And, um, one of the pieces we keep seeing are people wanting to push some of the applications into data centers that are near the, um, near these people. 'cause a lot of it is agriculture.

Some of the, uh, some of the precision ag agriculture work that's going on right now where, um, the turnaround time of being able to take a picture of a plant turnaround and tell whether it's a weed or not very quickly is very important. And if you have to ship it off from, you know, rural New Mexico to Chicago for that work to get done, the answer won't get back fast Enough. There's a whole melding together there of a, you know, John Deere.

Absolutely not, not the conventional tractor, but a a a robotic tractor and, and a whole bunch of lasers. So this is, this is the awesome part of, of what's going on out at some of the edges. It's, we're not gonna use chemicals to deal with weeds.

We're gonna zap them with laser. Shh guy, would you like to be zap with a laser today? Sure.

Just as soon as I identify whether that plant is a weed or not, you can zap, zap away. Um, I think, you know, the melting pot idea is a really good way to think of the edge because I think there are, you just described one to my mind one force that is, uh, uh, part of one half of the melting pot, which is the classic edge scenario of, um, uh, unusual conditions, extreme conditions, not under a desk, not in a data center, not cooled properly or normally, or any of that sort of thing on the one hand. And then on the other hand, um, what you didn't mention is, um, there's a lot more flexibility.

I, uh, that we see in platform architecture, um, form factor and so forth. Some of the form factor is required by the conditions. But, uh, the other thing is that there, there seem to be a lot more, uh, there's a lot more of an ability to build proprietary platforms.

Um, you don't have to worry as much, you know, just thinking of the processor, uh, as it being X 86 or arm or anything else. We've seen a lot more variety there. So you put those together and that is the melting pot for, uh, different looking even a, you know, a full rack cluster out in the field, um, might be significantly different from anything you'd seen in a data center.

And, uh, the vendors are finding ways to still allow those to interface back into the core. Absolutely. Yeah.

I actually don't, I agree and disagree with you. I wouldn't actually call them the edge melting pot at all. I think what we are basically doing is we're actually just shifting stuff.

We switch it from the data centers to the cloud, now we're shifting it to the edge. Could have changed in the form factor, but in the end we're still doing the same stuff. We're just doing it faster because it is closer to the source of the, where the data's initially corrected, collected.

So I wouldn't necess say it's a molding pot, I would just say we're just redistributing what we already have and finding different ways to manage it. Is that just part of the 50 year cycle of the, you know, centralized computing to distribute it and going back to centralized and we're getting distributed again? Well, they're all going over your networks.

So where's the problem? The problem has to do with latency, bandwidth, availability, that sort of thing. One of the things that, uh, I've always complained about mobile apps, for example, is that so many of them seem to be developed with this idea that they're, you're getting, you know, at least two bars on some, uh, local tower.

And the reality of it is that that's not the case even today. So there are a lot of shifting conditions that I think, uh, Ivan, uh, necessitate design changes in the systems themselves and that causes them to diver and be more, um, be more diverse in the edge. That is not the same thing to me at all of this shifting back and forth.

Well, I would be question is what piece of the edge is the most important? Is it the, is it the interconnection? Because if it is the interconnection, if you're talking about a cell provider, most of the time your handset is being back hauled to a major city anyway.

So if you're trying to rely on cellular systems, you're not getting the latency benefit of having things local and on, you know, at the edge. But are we seeing that there is the, again, the melting pot, this idea that the mobile devices are actually melding with the devices at the edge. And so you're getting something that has a mobile device architecture.

It'll often have a mobile devices CPU and may well have the offloads from mobile device, right? Modern mobile devices do AI locally. Uh, yet we're trying to deliver things that we used to deliver in the cloud to, to Ivan's point, we're bringing these things together, but having to make changes to them as well.

Because of course, you throw things into a melting pot, they lose their own shape and they assume the shape of the pot. And the problem we're trying to solve here is not one of, we've got a very small number of huge clusters. And this is absolutely a vast story, is you're not managing a, a single huge cluster or small number of huge clusters.

And so the tools that are good for managing a single huge cluster are terrible for managing a massively distributed collection of things that probably don't even need to be clusters. And so we definitely see this, there's a different set of requirements that are gonna drive a different set of solutions, but are we seeing the elements that make up these solutions getting melted together? Is the software and the hardware getting glued together more?

Is the connectivity and the user interfaces getting glued, which bits are melting together into this weird crowd, melting Them into the same pot? Just grinds me that, you know, through the years, Moore's law has given us infinite computing for all intents and purposes and infinite and infinite memory. Everything we've been doing has just been, you know, it, nobody cares about the individual cycles anymore because we've got so many of 'em.

They just throw things at that. So if we go further, is the edge gonna be into your phone? Are we going to start having people that we're dedicating parts of our phone to, um, you know, to somebody's use so that they can push that edge piece to us when we, when we subscribe to their services?

You see in the end, the edge is totally dependent upon the network. So as far as I'm concerned, the more you push things out and pressures on the edge, the more important the network becomes. And we're even seeing a lot of applications that were in the cloud being taken back into the data center in order to power things at the edge.

So therefore the network becomes the key thing. And in the end, every single edge device, edge device purely becomes an endpoint delivering certain capabilities, which in the end is dependent upon the network to be able to present the big picture for the organization. And yet we want our applications to still run at that edge when a snowstorm comes in and takes out all of our fixed line networking, uh, and all of the cell networks are then overloaded by people trying to contact one another, all of those cell sites go down 'cause they're covered in ice.

And so one of the characteristics we see at edge though, is the need for autonomous disconnected operation. Survivability. Survivability.

So not a dependence on something that's in the cloud in the data center for selling another gallon of gas because, you know, you've gotta fill up your generator at this time. Hopefully you've got your generator filled earlier, filled early, uh, but if you've got no diesel, no petrol left in your, um, in your generator, you need to be able to buy petrol in the middle of the snowstorm. And if it relies on the network being up, Well, that's why I really enjoyed Chris's illustration this morning about his child, when his child was at home.

You know, he used to have to make sure it's, he, he brushed his teeth before he went to school, had all his books packed, that sort of thing. Now, as a university, it's a case if you don't brush your teeth, you won't get the girlfriend. If you don't actually do your homework, you won't pass your, pass the syllabus.

Right? And then, so therefore we sort of move stuff out of the data center to the edge. Yes, survivability is important, but in the end that survivability is only there for a limited period of time and then the connectivity has to resume in order for us to be able to have something that really works.

But to me, what necessitates this melting pot approach? Well, let me ask, uh, you gen a question. The applications, the, you know, categorized as edge applications, um, would you say that they are, uh, uh, more stringent in their requirements ultimately, um, in order to run properly compared to any applications or generally applications that can just run in the cloud or run in the data center?

Are there more requirements for edge applications? I think it very much depends upon the sector, right? So even you look at, uh, industry, you look at manufacturing, uh, you look at 24 hour manufacturing for instance.

Yes, there has to be stringent requirements for those things, but that's not applicable for all things. For instance, you go back to, uh, the illustration about, Um, Having a look at crops. Okay, the latency requirements for that is a lot less than a manufacturing plant.

And so therefore all depends upon exactly what the application is and the sector that's using It. My contention is that, um, with the additional requirements, the physical requirements, security requirements, connectivity requirements, um, some of which may be low, but there's still requirements, you sort of having this piling on of requirements, that's what drives the melting pot. But I guess what you're saying, Ivan, is that it does drive the melting pot in some cases, but it does not drive any kind of melting pot in others.

So there you have the answer to the question, both yes and no. This is a perfect podcast. So We're returning to the consultant's answer of yes, but it depends.

Yeah, very good, very good. I I, I just observe this, uh, uh, diversity of platform, both software platform and hardware platform, uh, including networking, um, that I don't feel exists as much in either the cloud or the data center where there's a certain drive towards uniformity, um, just for management purposes, management and operation purposes. Um, there's the same drive towards uniformity in an edge application.

It's just that the limits of that are, are, uh, uh, greater because of those requirements. But I guess to your point, that doesn't apply in, in, you know, in as generally as I thought. Well, That's one of the, one of the vendors that presented today.

That's why I really enjoyed what they presented, because it allows you to be able to come up with a certain level of uniformity based on containerization and push that out to the edge and manage it. The, the, the, it comes from the containerization and then that's a hurdle for you to go over. But once you've gone there, that's not a melting pot.

I would agree. So one of the things that the vendors that's presenting for us tomorrow, our time of course last week, your time is Zita and they're definitely a, an organization where there's this real tight coupling between the hardware that's deployed out and the the actual hardware configuration and the applications that are being delivered on top of them. So I think tomorrow we're gonna get more of that story around how they melded together these different pieces and how interdependent at these edge locations.

But is that what we're really trying to look for? I mean, we have general purpose CPUs, we should be using them for general purposes. The any, any CPU should be able to run any of these, uh, containers that we want to run.

You know, I don't care if it's an arm, I don't care if it's, you know, risk five, I don't care if it's X 86, it ought to be able to run these things for them. And we ought to just have, you know, generic devices that maybe we give them part of our phone. Like I said earlier, maybe, maybe it's a piece of, you know, somebody's car when you're driving around that you've got some excess computing capacity.

I know that's, that's the world of the edge. I mean, what, and that's a melting pot more than anything I've ever seen in my life. The idea of just using whatever, um, resources are available at the time, It'd be amazing to, to get a, a faster response to the judgment of whether that's a weed or not, because somebody is driving a Tesla past your field at the time.

Well that's, that, that's also one of the things I appreciate about ssa. They had the ability to be able to do discovery of all the devices on the network and based on the configuration of that device, determine which edge applications could be deployed on that device. And I really think that's what's we're looking for in the end.

Because in the end, we don't wanna have to go and deploy lots of new edge devices. If we've got co server sitting out there, let's reuse them, right? Don't bring in something new when what's old will do well enough.

And so therefore, by actually having that discovery capability and then backtracking it to exactly what can be delivered on it is a huge advantage to organizations that are distributed. So you're looking for a converged platform that is allowing us to just deliver applications without so much of a, um, dictation of what's actually in the, in the, uh, remote sites. I recall Steven talking about FoST, um, Steven FoST, our founder, talking about the days before commodity hardware where every application that was being deployed out to a retail location had its own server and its own connection.

Those things have definitely been melted together into this commodity platform for running your applications. And we're definitely seeing more containerization is another way of melting them together. That, that to me is really the pendulum swing.

I mean, I think the grand pendulum swing in this industry is betwe is more or less between infrastructure and application. The center of gravity for, um, the center of gravity for well application. Sp uh, uh, I guess the center of gravity for the stack, the center of gravity for decision making is the, uh, dev application software developer, um, uh, adapting to the new capabilities of the platform or the platforms enabling, uh, new software models.

That's what containerization is supposed to do. Um, and maybe we're just witnessing the same thing distributed farther out to the edge, But also you think about is the application overlay, which in the end is what runs the business. It's the value to the business.

So the network, for instance, underlay becomes this underlay, right? But if we can actually manage and track the, uh, the application overlay and that determines exactly how the, the business is gonna perform based on that value presented by it so that everything else really becomes commodity in the end. Grace Hopper had a, um, or talked to the NSA in the early eighties and she was talking about how things were gonna be moving into systems of computers at the time, and she said that, you know, you know, at the time they had a big mainframe and you had, um, you know, had you had the processing power going on it, and, and the engineering was on it and everybody else was on it and, and payroll was on it.

But as we got cheaper and cheaper computers, they started getting their own systems and so they could actually break 'em apart. And the idea was now that things are, you know, the, the price model is getting to the point where we can break these apart and make them work better on their own, everybody was using their own computer and they got to design them the way they needed them to, but then they realized that all the connections that you have to go back and forth, so the underlying underlying pieces of the networks talking to each other and the ability of tying all these things together is what's gonna be key in all of these. It always has been.

And I think the containerization gives us the ability to, you know, first we had virtual virtualization and now the containerization we're gonna able, we're gonna be able to use bigger, bigger computers and have multiple applications running continuously that are all gonna need that separation. Yeah. And that allows us to be able to manage and optimize the application path as opposed to the network path.

And so therefore, in the end, it's, everything's about the application. Everything we do is gonna be making sure that we have the application and the, the underlying internet infrastructure that can give us the latency and the performance that we need. It's all in support of the business need that we have at the end.

And it's gonna have to be a melting pot for everything to, you know, boil down and get to the point where it all helps. I do think that there, that puts a certain new level of stress or, uh, a burden responsibility on the application designers, developers, and managers. And so that is its own kind of learning cycle and learning curve.

That was one I ran into today. We hit, we saw people that were able to orchestrate containers, but some of the questions they're gonna have in the long run are gonna be whether or not the container is enough. I mean, the, if the, if the actual application can't deal with the idea of part of it falling away And part of the issue, You know, These are somewhat esoteric application development skills.

Um, they, they, they are extremely valuable, not necessarily fully recognized until teams have gotten pretty far down a development path. And so, you know, that's where you can see that's where you have that kind of experience I just described, uh, about using my phone and there's just not enough connectivity. And on the app I'm working with doesn't work that you could call that poor application development, but it's just an area that, you know, needs to continue to mature.

If it's being given this power, it needs to take that power seriously. Maybe I can ask you guys a question based on that. Do you see the focus of vendors shifting on selling, selling to the IT folks, to selling to the application folks?

Because in the end, the application is king. I think that's the swing I was referring to earlier. I feel very much, um, this, uh, this weight in the last maybe two years, um, more on the side of the application developers and I, I I credit and blame containerization for this.

It was, you know, the supposed easy button so that the application developers can just define their environments and just go ahead and, uh, you know, create their microservices based applications. And the infrastructure folks, uh, the interface between the two is the in infrastructure folks are able to supply the service levels required for that, and they don't have to have this continuous back and forth. So who should hold the purse strings, the application guys?

Should the application guys be controlling the budget for the infrastructure? Or are we in a sort of service provider place? Is this where we're starting to come?

You set me up. Of course. That's, you know, that's where the rubber hits the road, isn't it?

Because, uh, uh, the infrastructure guys should control their infrastructure budget. And I think one of the things we heard today, um, and I think it was from SSA, was, uh, that they are, uh, uh, in their own work, they're empowering, um, as they build, you know, their business. They are empowering the infrastructure folks to speak to the needs of the application developers and managers.

So the, so Alice's point then is that the infrastructure guys should be the service provider for the application guys who are then service providers for the business folks. Well, okay, so I should say that's the vision. That's, and I hear these discussions all the time from customers who are very into containerization and into furthering modernization through containerization that they seem to take this as an easy button and it really isn't.

So I, I, I'd like to drag us back from the precipice of, of whether DevOps and, um, platform engineering are important things. 'cause I, I can see that the, the whole melding together, the melting together of these different technologies in here, we were heading for this idea that the technologies are coming together in this edge location and that a whole bunch of different technologies are being brought together and used in ways that they weren't previously used. That's, that's where I'm coming from, from this.

Not Only are they being used that way, they can't be used without each other. The interaction is, is the key to all of It. But I need, as we're coming to the end of the, the show, I'd like each of you to just bring some final thoughts on this idea that the edge is, is a melting pot.

And I'll start with Guy because he, he clearly has something to say. Well, I think, uh, Ivan has demonstrated to me pretty well that, um, the scope of that melting pot is necessarily limited. In fact, when I'm thinking about these additional requirements, um, which I do believe exist in, you know, almost all edge cases, essentially, um, many of those requirements lead more towards, let's say, purity than towards a melting pot approach.

So I do think that it's the consultant's answer, which is that it depends, but importantly to this discussion, it depends on the application. You need to understand the application it's requirements, what outcomes you're looking for that classic chain of value in order to know whether the melting pot approach is the right approach or the right strategy from your vendors. And I don't think it's a melting pot at the edge.

I think it's the melting pot that empowers the edge. Because in the end, yes, we want to deliver from the edge, but everything else we have to make sure fits in with what we're looking for in business perspective. Is that your final word for me?

That's my final comment. Excellent. In that case, John, surprisingly, you get to have the final wood from the Pen.

Final word would be, I think it absolutely is true. It's a melting pot. We have, we have applications that have to run on top of equipment that has to talk to each other.

And if they aren't all working in harmony and in this melting pot together, they're not gonna be able to take care of the end user that we really need. And whether that is completely at the edge or sitting at a data center, it's gonna be a hodgepodge of all of those pieces. So the edge has to melt together.

And I think that the melting pot has always been here with our infrastructure and the applications that reside on top. I think one of the things that I've brought out really strongly is we've gotta think about what we're doing for the business. Ultimately, our whole purpose as the Infrastructure IT application developers team is to support the business to be more successful as a business.

And I think very importantly, that understanding of your contribution, how you can bring together your expertise and your knowledge and the technologies that are available to you to provide those solutions for businesses at the end. Thank you for joining us at the Tech Field Day podcast. Before we go, where can people continue this conversation?

I know we are going to continue this conversation later on at a social event, but if people want to contact you guy, where can they find you to carry the conversation off? com, but I can be found on Blue Sky as, uh, guy Courier's feed and also on x at the same address. Sorry, was that Guy Courier's?

Feed, Feed, feed, like information feed. I apologize to the viewers. John, where can we find this that you to carry on?

I'm a little rather old fashioned. net email is probably the easiest way to get ahold of me. And for me, the best way is to go on LinkedIn and search for Ivan McPhee, M-C-P-H-E-E.

Of course, you can always find me as Alistair Cook or a Cook or deas, NZ, uh, just if you're searching for a cook there, uh, Alistair Cook, there are a couple of other guys by that name. But, uh, put some technology in, you'll find me. So thank you for listening to this episode of the Tick Field Day podcast.

If you enjoyed the discussion, please subscribe to our YouTube channel or use your favorite podcast application as, uh, so you don't miss a single episode. Also, do consider giving us a review and a like that helps a lot for other people to be able to find this awesome content. This podcast was brought to you by the Tech Field Day, home of IT experts from across the enterprise, part of the Futurum Group.

com/podcasts or view us on our wonderful tech strong TV platform. Thanks for listening and we'll see you next week.