HS070: Living on the Edge: The Present Future of Edge Architecture

Right now we’re in the first generation of edge architecture. We won’t even really know how to define it until we’re past it. Greg and Johna discuss the operational milieu in which the edge is forming: Vast numbers of IoT devices, increased remote computing capabilities, questions around cloud operational costs and efficiency, and vendors jostling for sweet spots in the emerging market. Johna emphasizes the need for businesses to put together workload placement plans. Greg shares his struggles with integrating his home lighting smart devices.

Episode Transcript

This episode was transcribed by AI and lightly formatted. We make these transcripts available to help with content accessibility and searchability but we can’t guarantee accuracy. There are likely to be errors and inaccuracies in the transcription.

Johna Till Johnson (00:00:00) – Welcome to Heavy Strategy. The technology strategy show about unanswered questions, not unquestioned answers. Greg and I are going to be talking today about the new edge architecture, which is also known as when cloud is not the answer. What was the question again? and and it’s an unanswered question.

Greg Ferro (00:00:17) – I love the way you said new edge architectures, if we know what it is, because I don’t think I would probably have said emerging or what we think it’s going to be, because I don’t think the edge architectures that we have are in any way coalesced into some sort of coherent like in the technology lifecycle. What we tend to see is things we can see the future. We know that edge is going to be a thing, but what we also know is that it’s going to take another 5 to 10 years to crystallize into a particular direction. Does that make sense?

Johna Till Johnson (00:00:45) – It does. I would I would push back that it’s 5 to 10 years because the paces are always accelerating. You know, it’s all a second derivative issue.

Johna Till Johnson (00:00:52) – Things are getting faster, faster. but but I do agree, it is true that when we say the new edge architecture that sounds like an unquestioned answer as opposed to an unanswered question. And I think it is absolutely true that the single thing we know about the new edge architectures, that we don’t know what it is.

Greg Ferro (00:01:10) – Know, and it’s it’s emerging and it’s certainly going to evolve, because as we build the first generation of edge products, that’s going to inform the second generation I and my gut feel is, is that the first generation will be radically different to what we see over time. So, you know, what we call edge today might be in manufacturing, which is just sensors. Whereas now people are talking about can we actually put machine learning at the edge and build the models at the edge of the network in real time, by the way, whereas the general assumption around AI has always been we have to export all the data to the cloud where there’s enough compute capacity to do AI processing to build the models.

Greg Ferro (00:01:48) – So I think there’s a lot of evolution to go on.

Johna Till Johnson (00:01:51) – In fairness, three years ago we put together an IoT infrastructure architecture that talks about exactly what you’re talking about. And one of the things we guide our clients to is this notion that you’ve got sensors that are, you know, both actuators and sensors because actuators take action, sensors gather information. Both of them are feeding to an edge device and edge computing device that’s living at the edge of the actual manufacturing facility, which may then be connecting to a cloud edge device, which may then be connecting back to the cloud. So there may be multiple tiers in this edge architecture. It’s true that we might have missed a few nuances here, but philosophically I agree completely. The idea of an on premise box that’s that’s doing a lot of the AI processing and is communicating back to other to a distributed network of other boxes is kind of the way people should probably start thinking.

Greg Ferro (00:02:40) – Yeah. So I think what you’re alluding to there is the thing that we’ve seen bloviating about extensively.

Greg Ferro (00:02:45) – I love that word bloviating.

Johna Till Johnson (00:02:47) – Which is true connected IoT.

Greg Ferro (00:02:49) – The idea that a lot of small, cheap, low to medium cost devices will be enabled to communicate across a network to something.

Johna Till Johnson (00:03:00) – Yeah. To to a to multiple tiers of some things. As is the is the magic magic piece that I would inject. Because, you know, one of the challenges with IoT generally is that most of the architectures were developed by data specialists, and they are very, very worried about how to get database A and a database B. And we can talk about streaming databases and real time databases. And there’s lots and lots of interesting technology about that. And I could talk your ear off about it, but it’s fundamentally not the problem because the problem is we have very, very different data streams with very different requirements for bandwidth and latency that have to hit pre-processing or processing at particular points in the, in the infrastructure. And that’s where edge comes in.

Greg Ferro (00:03:45) – Yeah, I think and the interesting part about this is going to be obviously the key part here is connectivity.

Greg Ferro (00:03:48) – That is the network where these devices connect and talk to something. And I think the original idea was that everything would connect at the edge and then just stream data into the cloud, and it’s pretty quickly become clear that’s not going to scale because the yeah, you know, the idea of a factory with, you know, five, 5000 sensors or 10,000 sensors or 20,000 or whatever, whatever ridiculous number you want to throw out, which is not unrealistic. Like, you know, we’re talking now about paying cents to attach a vibration sensor to a machine. Well, that’s very useful information, because it tells you that the device can be coming out of alignment or that the engine is wearing and needs service or whatever, but it can’t actually do anything intelligent. You have to stream the vibration data somewhere. And I think the original idea that people had was I would just stream it all to the cloud. That would make it really easy. And then people realize that, hang on 5000 devices, streaming data to the cloud, not very sustainable, you know, 50,000.

Johna Till Johnson (00:04:43) – Or 500,000 or 1 million. Yes. Yeah. And I but the one thing that, you know, because we’re old curmudgeons, always bubbles to the top is didn’t we have this conversation already? And I do remember back in the 90s, my very first, report that I wrote. Analyst long before it became a chief technology officer was exactly about this issue, which is, you know, are you where do you where do you put the computing? Do you put it, do you put it in some centralized data center in the middle of nowhere, i.e. the cloud? Or do you put it very close to where the data is coming from, i.e. the edge? And it turns out that with the new edge architecture, whatever it turns out to be, the emerging edge architecture, the answer is, as always, both. Which brings which brings another question to the top, which, you know, Greg, you’ve been arguing this for a while. How is this not just the pendulum swinging back to, you know, centralized data centers move the centralized data centers to the cloud? Now let’s swing the pendulum all the way back to put the data centers at the edge and call them edge computing.

Johna Till Johnson (00:05:48) – And, you know, so.

Greg Ferro (00:05:49) – I think the first thing that you have to keep in mind is that there’s a lot of marketing money attached to off prem cloud or, you know, the big cloud providers have managed to create a lot of hype around offering a new way to run technology. Right? You use it, you rent it, and it taps into what I believe is the the lack of transparency. We don’t know how much compute. If I write a program, you know, whether in whatever language you like, how much resource do I need? Well, I actually don’t know. So how many computers do I need to buy to run that. And that’s an uncertainty that the cloud tends to address. Now that’s just one angle of why a public cloud is useful. It also lets you take risks because you don’t have to, you know, have a large capital spend and all that sort of stuff. But to me, the original idea was that the cloud is the place to do this.

Greg Ferro (00:06:36) – It’s also got advantages in operations. It’s smoother, lower overheads and so forth, but equally it’s quite expensive. I think the general consensus now is coming around to yes, the public cloud is cheap. If you don’t have to think of certain things, like you don’t have to have operational stuff, but it’s expensive if you figure that actually doing all that yourself is significantly cheaper and there’s less business risk if it’s operated well, you know, you can operate a car badly and it’ll break down all the time, or you can operate and own and operate a car quite well, as long as you service it regularly and understand what you own. Now there’s so there’s two schools of thought there.

Johna Till Johnson (00:07:12) – Yeah, I’ll just interrupt here because it’s important, the schools of thought. And then there’s also data. And we went out and capture the data. And it turns out that both schools of thought are correct. Again, if you operate at poorly, you are. You’re paying a premium of upwards of 20% over doing it yourself.

Johna Till Johnson (00:07:28) – If you operate it well, you can actually be saving up to up to. The maximum amount you can save is about 40%, but that really requires operating well with a whole lot of details that are above and beyond the scope of this conversation. That’s right. Yeah, but we do have the data on what it costs to operate your cloud infrastructure.

Greg Ferro (00:08:26) – Yeah owning a business. Operating a business requires good operational control, and a lot of businesses actually don’t. They just focus on a few of them and hope to make a profit. And that’s sort of something that we always do miss, you know? And also being technology centric is a key aspect of that. But that said, I mean, the edge architecture is going to be something at the edge of the network streaming some sort of information back. And I think we’re going to end up with a tiered… I think you alluded to this in the early stage. There’s going to be on prem gateways of some sort. We’re going to see data streaming off devices and then aggregated or collected and then shipped off to the cloud would be the first phase. Potentially in a second phase would be we would actually see the edge device or that hub device or that data collector device become a machine learning point. That is, it’s actually got models loaded into it. And as the data streams in the machine learning models on there start to detect anomalies or signals, or.

Johna Till Johnson (00:09:21) – I would agree with that. With the models at the edge, the only thing I would say is that’s actually already happening. a lot of the cloud providers actually provide edge boxes, and customers have been putting those models in there. So while we’re saying that’s future, that’s actually rearview mirror stuff. Yeah. the one thing.

Greg Ferro (00:09:38) – What I’m saying here is that we’re actually talking about, say, you know, a software VM or a container. Yeah, that’s right.

Johna Till Johnson (00:09:43) – But that’s what I’m saying. People are already doing that. My clients are already doing that. That’s anyone with a manufacturing floor is going, yeah, Greg. Joanna. That’s nice. We did that three years ago. What’s next? But the thing I want…

Greg Ferro (00:09:54) – Many of the people listening aren’t right.

Johna Till Johnson (00:09:57) – The absolutely true, in this case, the manufacturing folks are bleeding edge here, and the rest of us are going, What? What an idea. But the capability, the capability is here is the important thing.

Johna Till Johnson (00:10:07) – It’s not like we need the capability to emerge. Apologies for that. It’s not like we need the capability to emerge. It’s here. But I want to sort of circle around and raise this question because you’re talking about two tiers, that there’s a carrier edge, sorry, that there’s a premise edge and a cloud edge. What I’m seeing is everybody wants to get into the game. So the carriers talk about carrier edge computing, and the cloud providers talk about cloud edge computing and the, you know, folks that sell direct to enterprises say we have enterprise edge computing. That would give us at least three tiers and probably a few more. If somebody can figure out how to make a few bucks from it. Do you think any of these tiers are actually, you know, must have and the other ones are nice to have.

Greg Ferro (00:10:51) – So I think we’ll see a wide diversity of solutions. Different applications will have application edges. Some of them will have different networking types. I think you also you’re looking at market evolution.

Greg Ferro (00:11:02) – So if you want to say telcos are well positioned to offer managed IoT services, they’re going to try. It doesn’t mean that it’s where you want to be. I don’t know if they’re going to.

Johna Till Johnson (00:11:13) – I’m not I’m not sure they’re going to offer managed IoT services, but they might offer the platform for those data analytics. Living in the carrier, in the carrier facility closer to you. So your choice then becomes, do I want to go all the way to the Amazon cloud? Do I want to have a box on my premise, or do I want to put my VM on Verizon or AT&T? You know, and I think that becomes a legitimate question.

Greg Ferro (00:11:36) – Yeah, I don’t I don’t know if the technology is ready to handle, you know, placing VMs or containers in, local infrastructure of inconsistent quality. So the advantage of doing, you know, containers and VMs in AWS or Azure or Google is that you get some sense of what is a consistent performance, you know, what the parameters of operation are, but the ability to run a VM on a 5G pop close to the edge because you want low latency.

Greg Ferro (00:12:03) – That seems to me to be stretching what the technology can achieve.

Johna Till Johnson (00:12:06) – I would disagree, I would disagree for the very simple reason that you know where I was hoping you would go when I asked punted you. The easy question is the difference between edge computing and your grandfather’s on prem computing is standardization, and one of the things the carriers are good at, and there aren’t very many, you know, sorry guys, but there really aren’t. No one is scale. And the other one is standardization. And occasionally they screw up. But, you know, they’re at least as good as the cloud providers at standardization. So you can be you know, it’s like McDonald’s. You walk in, the Big Mac always tastes like a Big Mac. You know, you go into that 5G tower, it’s going to look, smell and feel like every other 5G tower that Verizon or AT&T or or name your big carrier is operating. Now, if you decide to go into a model that says, I’ll go with 50 different carriers because, you know, that’s the way I want to do things for either you’re a global organization or you’re just weird, then yeah, you’re going to have the problem that you raised, Greg.

Johna Till Johnson (00:13:03) – And and I’m not sure that carrier edge is the right answer for you. Yeah. maybe it is.

Greg Ferro (00:13:08) – I don’t know, maybe I find it difficult to believe that, you know, 200 telcos around the world.

Johna Till Johnson (00:13:13) – Oh, yeah. That’s never happening, right? Yes.

Greg Ferro (00:13:16) – And even even some of them will be able to give me a VM in their infrastructure that would be able to perform at a consistent, predictable level. And then I think they.

Johna Till Johnson (00:13:26) – Can do that, that part, they that part they can do or at least they can give you the platform for you to put the VM. Because remember, they’re not managing the VM. They just need bare metal. You go, you go run the rest. that just means they have to have consistent power, consistent backup, consistent, you know, consistent CPUs. I think even a carrier can handle that. But the magic here is that you have to have the same carrier. It can’t be, you know, carrier X and carrier Y and carriers.

Greg Ferro (00:13:51) – Yeah. And the problem with carriers is that they’re not global. They’re right. Hyperlocal in a sense. You know. So for example, in the UK, the UK telcos are just in the UK. So how are you going to spawn in. And you know, it’s like people who told me that self-driving cars need low latency 5G. Well, that makes no sense. If I drive from the UK to France, how does my car suddenly connect to the 5G network and work properly?

Johna Till Johnson (00:14:14) – Well, see see here in the United States, we don’t actually have that problem because we don’t drive from New York to London because we get wet.

Greg Ferro (00:14:22) – No, but even if you drive from state to state, you would have to go from Vodafone to AT&T to T-Mobile.

Johna Till Johnson (00:14:27) – Yes, but but that’s just weirdness of you guys over in Europe. You know, all that said, we really suck in the United States at standardization, but that’s a whole different question. But coming back to it, yes, I think there’s a question mark around the carrier edge.

Johna Till Johnson (00:14:41) – That said, the carriers are highly motivated to get your dollars or euros or pounds. So I expect that will emerge. And as you said, you know, maybe there use cases for it.

Greg Ferro (00:14:51) – I think the existing colo companies are more likely because they have much more culture and capability to build. And host VMs then.

Johna Till Johnson (00:15:01) – And container clusters, you know, see, that’s where I disagree with you. I think you may have missed what the carriers have been, have been up to for the past 20 years. They didn’t win the business in the end, but it wasn’t because of a failure of technology. It was a failure of sales. They just.

Greg Ferro (00:15:15) – It’s going to be super hard to convince me that after being mistreated as a telco customer for the better part of 35 years, that suddenly they’re going to get it right, I don’t know, but they’re.

Johna Till Johnson (00:15:24) – You were mistreated by the sales guys who didn’t understand what they were selling.

Greg Ferro (00:15:28) – By the engineers, by the executives, by the I’ve been inside, outside customer, victim of telcos.

Greg Ferro (00:15:34) – I don’t believe that their internal culture can be turned around. I think all of those organizations just need to turn into electricity generators.

Johna Till Johnson (00:15:41) – And yeah, I was actually talking to I was talking to a friend who was a former thought leader at a major telco, and he agrees with you. So there’s that. I mean, he really made the point. And what’s interesting is he came from a culture that, you know, he was before he joined the telco, he was his last gig was building data centers for very large companies. you know, like a global bank, for example. And his job was to change the culture inside the telco. And it worked for probably 6 or 7 years. And then the telco was like, yeah, never mind. That was just window dressing.

Greg Ferro (00:16:12) – Yeah.

Johna Till Johnson (00:16:13) – That inherent selling circuits.

Greg Ferro (00:16:14) – Yeah they do the circuits. It always snaps back to that. And it’s it’s just that’s the culture. You know unless you can you would have to actually change the accounting system. Their whole method of general ledger would have to be structurally changed to make that work well.

Johna Till Johnson (00:16:27) – Maybe not quite that badly, but you definitely have to change how you bonus salespeople. And that’s a much bigger problem that we’re not going to solve. So I think one of the, you know, one of the unanswered questions that is very important is how many edges are you going to have in your in your edge architecture? And you can ask Greg, you’ll get one, one answer. You can ask me, you’ll get another answer. And then there’s the real answer which will unfold itself over time. So so I think there’s there’s that. Are there other major unanswered questions that people should be thinking about? Because I’ve got one. But I’m curious if you have.

Greg Ferro (00:17:05) – I think the operational aspect can’t be underestimated. I think if you’re going to own, weather, like in my house, for example, if you take into account I’ve got some, I went through a phase of being enamored of home lighting. Right. Well, I now have 25 lights connected over, you know, whatever, and some smart plugs and smart radiators and blah, blah, blah, right.

Greg Ferro (00:17:27) – You know, none of them really interoperate. But the problem is, is that I use Apple Home to try and unify them all into a single thing. And it’s awful. So operating these things is really very difficult. And the lesson I take from, you know, doing small stuff and scaling up over the years is if you’re going to run a fleet of 10,000 edge devices, you need to have some sort of centralized zero trust. Yeah, you need tools, right? In fact, the IoT part, the little devices that go at the edge actually isn’t as important as… And that’s one of the lessons we learned from SD-WAN, from cloud computing. This idea of software operations that directly applies, in my view, to IoT and edge.

Johna Till Johnson (00:18:06) – Well, and I’m laughing because your example of managing the lights, you know, I have conversations with my friends around the household and they’ll talk about the latest gadget and, you know, oh, I got the next generation Roomba. And I’m like, yeah, I had a Roomba back in the 90s.

Johna Till Johnson (00:18:20) – now I have this thing called a broom. It’s got zero operational overhead. When the bristles wear out, I throw it out and get another one. And by the way, it’s biodegradable, so no impact.

Greg Ferro (00:18:29) – You get.

Johna Till Johnson (00:18:30) – That? If so. But but yes, I think the operational question of 10,000, 10,000 edge devices, whether they’re all yours or split among among you, the carrier and the cloud provider is a real kink in your ops model. I think the other thing that’s nearer term, and this is something that was raised to me by a number of several, enterprise organizations, is how the hell do you go convince upper management that now that we’ve switched all of our operations, quote unquote, to cloud, we have to actually bring some of them back to this thing that we haven’t defined yet called the edge. Like, how do you even build that argument? And that’s not an easy answer, because even though technologists look at this and say blah, blah, latency, blah, blah, lots of traffic streaming blah, blah, data models, I, you know, the the senior executives here, blah, blah, blah, blah, blah.

Johna Till Johnson (00:19:20) – And they just are like engineers, go back into your den.

Greg Ferro (00:19:23) – Yeah, I think so… I think ultimately the trick here is that a lot of these IoT services may end up as a SaaS. So you will buy a solution, not a bunch of IoT sensors and a management platform. And then you have to stitch it together yourself. I suspect you know, some of the incumbent players will be out there producing things, a solution, and that many companies or many businesses will end up saying, I’ve got some data stranded in this SaaS platform over here, and I’ve got another platform over here for IoT, SaaS, and I’ve got, you know, I’ve got one monitoring these machines, but I need it, I need a they’ve got an… The one over here from this machine maker, and I need to weld these two together. So I do feel that the unanswered, you know, issue there is if you buy it as SaaS, how do you stitch the data together to get something meaningful? Those SaaS platforms have to let you export the data into your own tool, or you’re then stuck building everything in someone else’s.

Greg Ferro (00:20:19) – So if you go to a telco and buy, I don’t know, some, you know, capacity, and then you buy an IoT platform and you start deploying, well, now you’ve got to I don’t know where you’re into who.

Johna Till Johnson (00:20:27) – Matched the yeah, who matched the capacity and the SLAs to the requirements. Which kind of brings us to the to two bigger issues, one of which and I just want to edge this one in here, pun intended. Is the notion of a workload placement plan. You need to know when your workloads, where your workloads need to go and figure out the characteristics of a workload that requires an edge and which flavor of edge. If you have multiple flavors of edge. And that should be built in because right now, most people’s workload placement plan is if it’s on prem, move it to the cloud, which is a false and be unworkable because not everything on prem should be moved to the cloud. And so you should this the emergence of edge drives the need for a workload placement plan.

Johna Till Johnson (00:21:13) – So that’s kind of you know that’s kind of thing number one. And thing number two is your operational point. Like if you’re architecting your network over here and you’ve got your architecting your edge strategy over there, and somebody has to marry these two together, guess what? All that operational savings you got by outsourcing everything, you’ve just turned all your architects and engineers into integrators. Not not designers. Which is fine, because that’s an important skill. But they’re still there and they’re still on your payroll and they’re still doing something incredibly important. So this idea that you just saved money, time, effort, energy by outsourcing is a canard?

Greg Ferro (00:21:50) – Well, it has always been a canard.

Johna Till Johnson (00:21:51) – As it always has been.

Greg Ferro (00:21:53) – But the secret with outsourcing is to make sure you’re gone by the time people realize that they’ve been duped. But that’s always been my theory.

Johna Till Johnson (00:22:00) – You know, I was I was actually talking to somebody who said, you know, I love the you guys have a healthy degree of cynicism on the show.

Johna Till Johnson (00:22:07) – You balance each other out. Sometimes you’re too optimistic and Greg is too cynical. I have to say that is a cynical, cynical as as death statement. But it’s actually true. It’s, you know, they’re just they’re just environments in which the right answer is, we shall outsource everything, boss. And then just set your, you know, set your alarm on your watch and start, you know, start looking for your next gig.

Greg Ferro (00:22:30) – I’ve been a victim of outsourcing in all dimensions. I’ve worked inside the outsourcer, worked inside the deal team. I’ve worked for customers who’ve been a victim of outsourcing and tried to dig them out of the hole. And it’s, you know, just it’s just. I do not believe that outsourcing ever worked. I am ultimately convinced that if you’re going to outsource something as SaaS, that’s still just outsourcing, all we’ve managed to do is bundle the outsource into a manageable contract that you can clearly delineate right? Where instead of saying, oh, manage my service, which is just an intangible fluff.

Greg Ferro (00:23:05) – And that just to make that relevant to what we talk about here, that’s what you have to do when you’re building an edge architecture is find ways to capture these into particular services or functions. And don’t say, oh, I’m going to buy in. You know, you can go out and buy third party boxes and attached sensors to them and then write software that does all of that. But you may well be better off served by buying that from somebody else because they’ve done the integration work for you.

Johna Till Johnson (00:23:30) – Yeah, absolutely. And actually all this talk of, you know, insourcing versus outsourcing reminds me of one of Robert Heinlein’s more memorable quotes, which goes something like this. you know, authoritarianism is based on the premise that one person knows more than many people. democracy is based on the assumption that many people know no more than one person. How does that work again? So basically, outsourcing is this idea of perennially chasing the grass is always greener. So after a while you decide your own team are idiots, and the only way to address that is to give it to the idiots on the other side of the fence.

Johna Till Johnson (00:24:07) – Yes. And then the new guy comes in, or gal comes in and says, oh, look, the the people on the other side of the fence were idiots. We can do this better ourselves and brings it back to our idiots.

Greg Ferro (00:24:15) – Now that sounds like capitalism and market competition. Exactly. It’s it’s again market. Competition only works when there’s a market that is multiple suppliers offering you similar products. and there you can quantify the competitive solution between them. For example, cars this car has, you know, they’re all functionally the same, but you may perceive value in the fact that this one’s got metallic paint, and this one only has, you know, a different type of engine. I don’t know, whatever it is that floats your boat around cars. I don’t get.

Johna Till Johnson (00:24:49) – No, you just what floats your boat is boats. But that’s a whole different conversation that some some day, Greg, we’re going to have to just have an entirely personal conversation on this podcast so people can people.

Johna Till Johnson (00:24:59) – You know what we talk about? Oh, they totally do. They totally do. There is life outside of technology, is all I can tell you.

Greg Ferro (00:25:06) – One thing about edge, though, that’s very interesting, is it’s going to have a dramatic impact on power, because I think as we see more of off grid, this idea of solar and batteries, distributed battery packs and this idea of shutting, you know, getting homes on multiple tariffs and shutting them down, and the same will apply to data centers. I suspect they all will want to say, I want to have a tariff for loads that I can shed in the event that there’s electricity shortages, if, you know, and this is linked to geopolitics to some extent. So it’s and it’s inevitable. But you know, we’ve seen governments across the world under invest in infrastructure up until, you know, for the last 50 years they’ve just been willing to say we need more power. We’ll just build more power stations here. Some billions go and build another coal here or build a nuclear over here.

Greg Ferro (00:25:48) – And the government are turning away from that and saying, well, we have to find ways to save electricity or to make electricity work better. And so we’re going to see IoT and edge play out in the home and in businesses. You know, if you’ve got an office, which is and you’re returning to work, but only 20% of your people are in the office, why are you running the air conditioning, food out, you know, at full beam?

Johna Till Johnson (00:26:08) – Well, and there’s also the question of why are you running the office? But again, off topic, let’s assume I agree. Yeah, I agree completely that this is your power infrastructure is something you’re going to want to very much rethink. And that’s one of the unanswered questions that’s incredibly important here. In fact, what’s interesting is an alternative energy company came to us and said, listen, we’re not sure what’s going on, but we have a use case. People are buying our stuff for data centers in California. We want to know what’s going on here.

Johna Till Johnson (00:26:39) – Is this an actually a trend? Is there a there there. And we were kind of like, oh, how interesting. And we got deeper into what they did and said, oh my gosh. Yes. You know, and it turned out it wasn’t really data centers. It was like little edge compute, you know, data center in a shipping container scenarios. And they’re, you know, their solution is actually kind of amazing because they can just bring their shipping container full of power, Park it next to your shipping container full of compute, and you’re off to the races. And you don’t need to worry about the grid, which we found fascinating. And then they did a total cost of ownership study that showed that this was actually cheaper in the long run than getting, you know, depending on what your your price per megawatt is for buying electricity, that their solution was cheaper. So so to add to that challenge of how do you operationally manage 10,000 edge devices is the next question is and they’re energy sources.

Greg Ferro (00:27:31) – And I think to me that’s probably the one edge thing that you could really talk to management about. Management will comprehend power consumption and costs. Everybody sort of understands power. And in the current well that’s not true. A lot of executives just don’t care. I was just going to say but.

Johna Till Johnson (00:27:47) – Yeah, I would say and there’s two problems with that. One is what you just were alluding to, which is not everybody understands power. It’s free as far as they’re concerned. And the other one is every time you go talking about something that’s outside your IT box, you get smacked on the head. Yeah. Oh, engineer, you don’t know anything about power. That’s our power people. And the power people are, you know, protecting their jobs. And sometimes, you know.

Greg Ferro (00:28:09) – People are the power people, too. Or.

Johna Till Johnson (00:28:11) – Oh, sometimes, but not for large organizations. you know, if you if you have a large on prem data center that you’ve been managing historically for 30 years.

Johna Till Johnson (00:28:20) – Yeah, probably you have conversations about this. But for folks that have successfully outsourced a crowd to cloud to and.

Greg Ferro (00:28:26) – Rent their premises source, they’re building ownership and so forth.

Johna Till Johnson (00:28:30) – Exactly. And for folks that are just newer and have come up, you know, their digital native organizations, DINO’s, those folks, I think there’s a real risk that you’re going to get smacked on the head and told to stay in your lane. Yeah. Because, yeah. Because in general, when it comes in and says, listen, you’re spending like $10 million on real estate and, you know, $5 million on power. And I’d like you to invest $1 million in it so we can cut those two costs in half. The usual answer, the smart answer is hello? Yes, let’s go do it. The usual answer is stay in your lane.

Greg Ferro (00:29:03) – In your lane. Yeah. That’s always well, that’s always a problem. But you know. Yeah, it is a challenge. And often it has been seen as a cost center, not a profit center.

Greg Ferro (00:29:10) – But if you can go to the CFO and say we could cut 20% from our electricity bill, provided you’re not outsourcing it to the building owner or, you know, a coder, because that’s what a lot of people do. We outsource, you know, data centers. We put our stuff in someone else’s data center. You know, if it’s in a colo. They’re the ones doing the power generation. But, you know, notwithstanding, there will be solutions around this, I think, over time. But that’s the one that I think would be more easily communicated. The idea that and.

Johna Till Johnson (00:29:38) – And I think at this point we have we have cleverly lined up a whole bunch of unanswered questions for folks to go explore on their own. we would love for you guys to give us feedback. as Greg likes to call it, follow up. if you how do they go? Give us a few. Greg.

Greg Ferro (00:29:54) – You can head on over to Packet Pusheres dot net slash FU. We don’t ask for any personal details.

Greg Ferro (00:29:58) – You can give your comments anonymously. We want to respect your privacy, and we won’t harass you or use your personal information for evil acts of marketing. There’s lots more podcasts on the Packet Pushers network. Go on over to Packet Pushers dot net and you can see a list of them all. And thanks very much for listening to us, Johna and myself. Don’t forget to go over to Johna’ site, Community.Nemertes.com. You can join up the community that’s there and join in some discussions between various people on the same topics that we talked about here. And as always, thanks for listening. We’ll be back in a couple of weeks.