Today’s episode features an interview between Matt Trifiro and Dean Bubley, Founder and Director of Disruptive Analysis In this interview, Dean gives us his unfiltered views on Edge, IoT, 5G, their intersection with the telecoms industry, and why the future of edge is going to be a lot messier than you think.
Today’s episode features an interview between Matt Trifiro and Dean Bubley, Founder and Director of Disruptive Analysis
Dean is an independent technology industry analyst, futurist, speaker and consultant with over 25 years of experience.
In this interview, Dean gives us his unfiltered views on Edge, IoT, 5G, their intersection with the telecoms industry, and why the future is going to be a lot messier than you think.
Key Quotes
“Everything is going to be a messy hybrid; deeply inelegant and based on sort of midterm pragmatism, messed up by acquisitions, with awkwardness around local authorities and this real mess of overlapping jurisdictions and property rights and incumbency. It's going to be messy.”
“My main perspective is that latency is perhaps not the be all and end all we thought it was initially. And that actually it’s things like data sovereignty, security and control by enterprise, and this idea of interconnection is more important than people thought it was a few years ago.”
“One thing I often have a conversation with people about on edge is trying to calibrate where they are on the scale. I often say that edge has maybe nine orders of magnitude. I've spoken to people who think of edge as a megawatt data center in a tier three city. Down to other people who think it's a milliwatt processor on a chip, on a sensor. And so I'm like, right, you've got nine orders of magnitude, all of which people think that's the edge.”
“A lot of my clients are in telecoms and I think they will end up being both sellers and buyers of edge compute. The telecoms industry has to wake up this idea that actually there’s a marketplace, and sometimes it makes sense to be on the buying side and sometimes it makes sense to be on the selling side.”
“As much as all telcos chest thump about how they're all building out their 5G networks, it's incredibly expensive and the CFO's office often has a very different opinion than the marketing team's office. What I see as a really big opportunity is shared infrastructure because the economics is so compelling.”
“The assumption in the mobile industry is that the edge goes in the network. I actually think that possibly the network goes in the edge, and actually what you might find is these localized data centers that have chunks of each of many different network operators infrastructure in a neutral colo. And so it almost turns the paradigm around.”
Sponsors
Over the Edge is brought to you by the generous sponsorship of Catchpoint, NetFoundry, Ori Industries, Packet, Seagate, Vapor IO, and Zenlayer.
The featured sponsor of this episode of Over the Edge is NetFoundry. What do IoT apps, edge compute and edge data centers have in common? They need simple, secure networking. Unfortunately, SD-WAN and VPN are square pegs in round holes. NetFoundry solves the headache, providing software-only, zero trust networking, embeddable in any device or app. Go to NetFoundry.io to learn more.
Links
[00:00:00] Matt: Hi, this is Matt Trifiro CMO of edge infrastructure company, vapor IO, and co-chair foundation, state of the edge project today. I'm here with Dean Bubley, founder and director of disruptive analysis and independent technology industry analysis consulting firm.
[00:18:01] We're going to talk about Dean's role as an industry analyst, often a contrary one and macro trends at edge computing, AI IOT, private networks, 5g, and everything else. Hey Dean, how are you doing today?
[00:18:11] Dean: Yeah, thanks, Matt.
[00:18:13]Matt: Yeah. So, I really been looking forward to this because you and I, you know, we started out sparring on Twitter, which is, well, it's, it's often fun, sometimes devolves into nonsense, which our arguments may have, but I certainly gained respect for you in our, our sparring. Cause you've, clearly done your homework. And, you know, I I've come to consider you part of the, you know, the edge.
[00:18:37] Oh gee, the original gangsters. because you know, you think two years ago, which is like, well, 2018, when I started the state of the ed report and I'd just been in the industry for about a year and a half, there were very few people that were talking accurately about edge computing. that's changed a lot.
[00:18:53]so I want to cover all these things, but before we get into edge computing and stuff, I have a burning question, which is [00:19:00] where did you get the name? Disruptive deem.
[00:19:02] Dean: Ah, well, disruptive analysis. I called my company when I set it up in 2002 and, and really, I, I decided to do my own show and I thought, well, what's my reputation. And at the time. I was fairly contrarian. And that was at the time, not that long after Clayton Christensen talked about disruptive innovations and so on.
[00:19:27] So it was still quite a fresh term. And I remember the first time I announced it, I was in a, I think it was, The analyst event run by Cisco. And, everyone knew that I'd, I'd worked at another role and then started my own company, but didn't know what it was called. And so I put my hand up and said, Oh yeah, this is Dean Bobby from disruptive analysis at half the room turned around and I either applauded it or I got a thumbs off of everyone nodding, you know?
[00:19:51] Yeah. That makes sense.
[00:19:53] Matt: It, it does. It does. It actually fits you really well. And, either the, the brand has, grown to conform to you, or [00:20:00] you've grown to conform the ground, but it's very, it's very consistent at this point. So I I'm really, I'm really a fan. so how did you even get into technology? I mean, what's, what's your background there?
[00:20:09] Dean: Well, to be honest, my original background in technology came from the fact that my mother used to program mainframes in Fortran and the 1960s. So I actually grew up,
[00:20:17] Matt: that is that is really
[00:20:18] Dean: I grew up at one point, I think we had a terminal for the one of the UK government department mainframes. We were at a terminal at home and at various points of various bits of computing gear around the house.
[00:20:30] I think from the age for about four, we had more computers, the people in the house. and so. Yeah, that turned into an interest in technology. And I didn't, I actually studied physics rather than computing. but then I ended up at my first job as a strategy consulting, focusing on telecoms networks and got dropped right into the deep end because I had various clients at the
[00:20:52] time.
[00:20:54] would have been around about 1992, I think.
[00:20:59] Matt: The [00:21:00] internet was mostly about email and academia.
[00:21:03] Dean: I remember doing a project and I think 1994 for a client that was an online service provider that wanted to compete with companies. And in doing the research for it, we actually came to our client and said, have you heard of this internet thing? There's this thing called the web. That's just been invented.
[00:21:19] I think you should, you should probably check it out. So I definitely recall that.
[00:21:24] Matt: Yeah. Yeah. That's great. That's great. And, where did you first. Intersect with edge computing.
[00:21:31] Dean: Hmm, that's an interesting one. I think pretty F I probably have been hearing about it for a while, whether it was cool edge computing or not. back that 10 years ago, and I was talking about local breakout, for, mobile networks and with small cells and the idea that you'd need to have some sort of core network function.
[00:21:49] Well, that's what we now think of as the edge, but it could be potentially integrated into a small cell. And then also, I think I had a couple of the early presentations from, ETSI, when they were setting up [00:22:00] the mech, project, as part of NF functional virtualization work. And I think, I think the first time I actually wrote about it in any real depth, it was probably just over three years ago, when, Amazon bought whole foods and I speculated that, you know, 700.
[00:22:17] Exactly. So 700 retail retail outlets. We had lots of FedEx, recency and flood, but might, might be a good, a good option.
[00:22:24] Matt: Yeah, that's great. And, you know, you've had some time to evolve. Your thinking about edge in the industry has certainly, caught up with a fraction of a type. what is your overarching perspective on edge now?
[00:22:38]Dean: my main perspective is that I think that latency is perhaps not the be all and end all we thought it was initially. And that actually gives things like data sovereignty, it's, security and control by enterprise. And this idea of interconnection. is more important than I think that people thought it was a few years ago.
[00:22:59] The other thing I [00:23:00] I'm seeing slow convergence of the sort of views of different edge worlds from the cloud and data center world and the telecoms world. And up to a point in the IOT and the device world, although I still find people. Talk across each other quite a lot. Yeah. The one thing I often, I have a conversation with people about, for the first time on edge, it is trying to calibrate where they are on the scale of when I say that edge often has maybe nine orders of magnitude.
[00:23:31] And I've spoken to people who think of edge as a megawatt data center in a tier three city. Down to other people who think it's a milliwatt processor on a chip, on a sensor. And so I'm like, right, you've got nine orders of magnitude, all of which people think, I think that's the edge.
[00:23:50] Matt: So, how do you, how do you navigate that?
[00:23:53] Dean: That's a good question. I suppose it depends on who I'm talking to at first, the first thing is asking that question and find out which order of magnitude or [00:24:00] magnitudes apply to a given conversation. And if I'm talking to a sense of vendor or a, a gateway vendor, that's very different to someone who's doing edge data centers.
[00:24:09]and it's different again, to someone in the mobile telecoms industry where there's often a percent of perception, that edge is. yeah, it belongs inside the five G network, for example.
[00:24:22] Matt: Yeah.
[00:24:26] I mean, there's a lot to unpack in all that we could spend the bulk of the hour talking about all this. So maybe, maybe let's let's tease apart some of those threads. so the first thing you said was maybe it's not latency that matters. And I want to add that, you know, latency, is often used as a proxy for, for.
[00:24:43] Too inclusive of jitter. And just for the benefit of our listeners, the whole all benefit of low latency edge computing is to have a discreet and predictable response time and latency and jitter, which is a function of network hops and distance affect that. But [00:25:00] your statement was that may not be the most important thing or the thing that most people talk about. So why, why are these other things important?
[00:25:05] Can we talk about each of those? You mentioned security and some other
[00:25:07] Dean: Sure. Well, first off on the, on the, the time and response time, what I'd say is that there's this a bit of a bifurcation that I'm saying, but for a lot of applications, there is a desire for less latency, but here maybe say talking about going from a hundred milliseconds down to 30 or maybe 20. Or you're talking microseconds.
[00:25:33] And again, this is orders of magnitude again. And so in a factory environment, it may well be microseconds that matter, or even in a carb that how fast you can, you can
[00:25:43] apply the brakes. Yeah. Yeah. And so, so then the question is that bit in the middle, the sort of one to 10 milliseconds, everyone gets so excited about, is that really the most important buckets?
[00:25:58] Yeah. If I'm going one to 10 [00:26:00] tens or a hundred, a hundred to a thousand or below that in terms of microseconds, where is the value. In those orders of magnitude. And I'm not convinced that that's is where the, the near term value for edge computers. Because I think a lot of application developers know that the users will have lots of different network connections.
[00:26:22]some will be on sending us or will be on wifi. So I won't be on fiber, multiple hops, different interconnection points. And so, as long as this year's latency is better than last year's latency. Almost a trajectory then, then that's all good. I don't see this sudden discontinuous jumps. So we're all using one millisecond round trip robots.
[00:26:41] And.
[00:26:42] Matt: Yeah, I think, I think that's true. I think that's really true too. And it's, it's, it's interesting that, you know, so the evolution of. you know, my, my journey, and understanding what's gonna happen first with edge, I've come to a very similar conclusion that latency is important and reducing latency is important, but you're right.
[00:26:58] This like these absolute [00:27:00] discrete latencies have very limited use cases. And a lot depends on that last mile network, which, you know, certainly in the wireless world is still being built out. you know, the modem handshake and on a, on an LT, the network is,
[00:27:09] is
[00:27:09] audio_only_16780290_Dean_Bubley: Yeah.
[00:27:11] Dean: and then you add in, yeah, it will happen indoors over someone else's indoor wireless or wifi. You've got, you got delays in the device. So you and I have been cited in the same article about video conferencing. I noticed. And, and that's yeah, we're on other side of the planet here. So frankly, main lane latencies is fixed.
[00:27:31] It's the speed of light between. Where I am in London and, and where you are. yeah. And any amount of local breakout might help here and there, but fundamentally it's the planet.
[00:27:43] Matt: Yeah, well, and, and the way that I tend to explain this to people that are coming at it fresh, just to say that the biggest change that I see in the next decade or what I call the third act. Of the internet is we're moving from a world where it's primarily humans talking to machines, or I guess in this case, humans talking to [00:28:00] humans over video conference, two machines talking to machines, and you're right.
[00:28:03] There are these orders of magnitude, you know, machines, you know, 100 milliseconds is glacial, right? I mean, I could do a lot, a lot, a lot, a lot of, floating point operations in a hundred milliseconds. so. Along my journey. One of the things that I've become, fascinated with is how important the network is.
[00:28:19] And you mentioned interconnection, and I'm wondering if you could, you could talk a little bit about interconnection.
[00:28:23] Dean: Yeah. I mean, internet connection is actually something I've been spending a lot of time on recently, with some work that I've been doing, looking at, internet hearing, but also in telecoms, interconnection as well. And there's multiple different interconnection points. So the, the, the voice. One video that we have going between us at the moment will almost certainly transit multiple different access networks, service, service providers, internet backbone, providers, pairing points.
[00:28:51] So that is a fundamental characteristic of most applications, which are not single network. I think that in [00:29:00] particular, A single network application. that might be where a carrier works with a particular application provider. So let's say, in the UK, BT has a contract to run the public safety.
[00:29:16] What's called the emergency services network. so in that case, it's their infrastructure pretty much end to end. Whereas at the moment we're having a conversation on your, on. Your access provider, I'm on my access provider. And we've got, who knows what in between us. And I imagine from a, from an edge compute point of view, imagine that we wanted to, we were, yeah, actually in the same city, but we wanted to pay some virtual reality game or, or something like that.
[00:29:43]the probability that you and I will both be on the same mobile network, even if it was five G would be 30% or low, you know, I would be on. Carry a, you be on carry a B. And then the question is whilst we might individually [00:30:00] have the polygon rendering for that game done on the edge in carrier A's network and carry a BS network, the game play, and we're.
[00:30:09] Hitting a ball between each other or whatever it is. we'll interconnect somewhere deep back in the network. unless there is some sort of local pairing or local interconnect or shared local edge, That to my mind is a real obstacle for a lot of the, the edge use cases that we talk about.
[00:30:27] There's no way that every connected car is going to be on the same network. you're going to have this issue of not even knowing what network you connect to. So if you bought a car, do you know what, what network is connected to? Probably not. If your, I don't know, you invite the elevator engineer into your building, to fix the elevator.
[00:30:46] And they've got an AR headset, what networks nobody knows. and so who's edge compute. How do you get to it? There's so many unknown. And unless you have that in your mind, and think of all the defense, interconnection [00:31:00] scenarios and multi networks scenarios that might occur, you've got a problem because all of the bits of the compute and certain workloads might be done locally.
[00:31:10] Other things, including where the analytics is done. They've got a machine vision algorithm running on a cloud somewhere. They've got data stored somewhere else. They've got security provider. Yeah. All of the microservices. They're going to be littered all around the sort of cloud scale. And again, then it's a case of ham.
[00:31:30] It's less about the physical distance and more about the minimizing, the number and to do that. they all need to meet somewhere convenient.
[00:31:40] Matt: Yeah. And there needs to be probably software based interconnection, to handle those, especially, you mentioned mobile things in motion. So, so an interesting topic that directly, you know, sexist is what you mentioned right at the beginning of the interview, which was, you know, 10 years ago, you wrote about local local breakout.
[00:31:55] And I think local breakout is still a relatively unknown concept [00:32:00] outside of the. The folks that, you know, spend time with the GSM and ETSI and all that. Can you help our audience understand what local breakout is? How the world of wireless works prior to local breakout
[00:32:13] and why it's so important for edge computing?
[00:32:15] Dean: Sure. Let's start off with, with something which works differently. Talk about wifi. If you're listening to this, you know, on a laptop or on a phone connected to wifi, your wifi will connect to your ISP, which will go to the local, either central office or exchange, you know, depending on which part of the world you're in and what you call it.
[00:32:36] And it will then be sent. As fast as possible to the nearest internet
[00:32:44] Matt: Which could be hundreds or thousands of miles
[00:32:46] Dean: it might be, but, but, but there's no, there's no incentive or very little incentive for a cable company or a. A fixed network operator to put your traffic through the transport and core [00:33:00] network and out the other side to the internet, they want to dump it to the internet as fast as possible. Apart from the house, it means it doesn't clog up their transport network.
[00:33:08] It's someone else's problem. And the same thing. If you're downloading movies on Netflix or data's coming the other way, Yeah, it's in your ISP interest to let net Netflix deal with that problem as a soon into the network as possible. So whether it's sort of local breakout or break it. Yeah. there's an incentive with wifi and home broadband to get it off the network onto someone else's network.
[00:33:32]now cellular. Historically, all of the data traffic has gone back into the core of the network, where there are various functions around policy, around charging, and then perhaps just one or two links between the cellular network and the public internet. And one or two pairing points varies a bit depending which country you're in, but it essentially means that in.
[00:33:56] Even parts of the U S your, your data tropical mobile might go all the [00:34:00] way from California to Virginia before hitting the public internet
[00:34:04] Matt: Yeah, I
[00:34:04] mean,
[00:34:04] Dean: yet.
[00:34:05] Matt: route on your mobile phone and it's pretty
[00:34:07] Dean: so the idea of local breakout. Yeah. Can you typically, if you've got something like a small cell and I really care about latency, can I do what wifi does, which is essentially dump the traffic to the internet straight away, and then you run this political problem.
[00:34:23]perhaps regulatory problem, that, that means if I'm going to dump the traffic from a small cell to the internet, as fast as possible, I still need to work out how to charge for it, which means I need to have exactly. So I need to have a bunch of sort of core network functions pushed out into the network down as close as possible to
[00:34:47] Matt: Yeah. We're historically, they've never been, they've all been centralized for economy reasons and convenience
[00:34:52] Dean: Well, and so you could argue that local breakout was the original mobile use case for edge compute, [00:35:00] because the only way that you would be able to break that out is to have that control function at the edge. And actually, I, I, I think if I recall this was one of the reasons for network function, virtualization, because you wouldn't want to have a hardware EPC at all of those breakout points.
[00:35:19] You'd ideally want to do it in software and you'd ideally want to do it at the edge. and I think that there's a separate thing here, which is. When you talk to mobile networks, they're very keen to position, edge computing and what they call them. Multi-access edge computing, originally mobile edge computing as a customer facing cloud edge service for developers or content providers.
[00:35:44] Whereas action, the number one use case of a telecom, a telecom operators edge. Is for its own internal network functions. And whether that's for the core network function that it's running the edge, or increasingly [00:36:00] now we know with the open radio network functions. So the anchor tenant of a telcos edge compute is the telco itself.
[00:36:09] And
[00:36:09] then if there's any left over, then they can think about something it's cost.
[00:36:12] Matt: Yeah. Yeah. Well, well, and actually, you know, it's interesting, cause that's sort of meta and you're right. Most many people don't understand that that one use of edge computing is actually to support the infrastructure for edge computing itself. and when you think about the, you know, the, the wireless industry, historically, you know, you go buy a bunch of appliances from Erickson or Nokia and you tend to buy the same.
[00:36:34] Brand Radiohead with the same brand, whatever. and that, that created these, you know, these, these incredible lock-ins and the, and inconvenience, right? I mean, inside every one of these appliances is probably an Intel server. That's capable of running many network functions, not just the one yeah. For that, that unit.
[00:36:50] And so the idea of pulling all that off. And putting it into software and running it on, you know, what you would think of traditional data, the data center servers, is kind of an obvious, [00:37:00] but, but has been time consuming approach for a whole lot of reasons. Some of them political.
[00:37:04] Dean: And so you've got the, the main use case is running the network. Then there's a couple of other, very sort of uncle network, adjacent functions. So there's obviously a content delivery network. So having a CDN node,
[00:37:18] Matt: Right. Well, and having it in Iraq right next to the virtualized network functions makes even more sense.
[00:37:25] Dean: And also, security functions as well. yeah, there's various, there's various things like, anti DDoS that you may want to have out of the edge. for instance, so, so again, that's sort of a, a network internal or a network adjacent function, but in terms of running a game or I don't know, a financial transaction application or a robot.
[00:37:46] I'm a controller for a factory that's in a way, secondary and tertiary. Even if the marketing people like to talk about it first.
[00:37:53] Matt: Right. Yeah. Well, and, and it's, it's also interesting that, you know, you started out with the, the [00:38:00] controversial statement and I appreciate the controversy of maybe it's latency, that doesn't matter. and I think you're right when you're looking at those other larger applications, we most think about latency.
[00:38:09] Probably isn't the first thing. It's probably network interconnection, but when you look at the telco network and it's, and it's function, virtualization, The discrete latency is really important. I mean, the, the, the separate link which splits the Radiohead and the base band unit has to be 75 microseconds in almost all cases.
[00:38:26] Sometimes, you know, you buy the same radio from the same vendor. You get the 20, but that's 15
[00:38:30] kilometers, which means you have to have something that looks like a data center, certainly racks of servers within 15 kilometers of every
[00:38:38] radio head, every remote radio
[00:38:40] Dean: Yeah, but you could equally argue that that's like saying normally you have
[00:38:48] Matt: Well, that's true, but you know, what's interesting at this because the, the environmental requirements for running racks of servers are very different than environmental requirements of running traditional, telco [00:39:00] appliance, you know, well, and for a lot of reasons, one of them is, is just, you know, efficiencies and cooling and things like that.
[00:39:06] And airflow and hot aisles and cold aisles things that, that the telcos have. Have not wide, widely adopted. the other thing is, and this is I'm actually really interested in your opinion on this. and I've formed this opinion from talking to a lot of, not only peers analysts, but also the telcos is that as much as all the telcos, you know, chest thump about how they're all building out their five G networks, it's incredibly expensive and, The CFO's office has a very different opinion than the marketing team's office and what I'm seeing, or what I see as a really big opportunity is shared infrastructure because the economics has shared infrastructure so compelling.
[00:39:46] If you have to build something that looks like a data center, because you want to put. Know, eight kilowatts of equipment in some location versus leasing a rack in someone else's data center in a similar, adjacent location, much more economic to do [00:40:00] that much more economic to use somebody else's fiber, much more economic potentially to use someone else's small cells.
[00:40:04] So how, how talk to me about, about shared infrastructure and telcos?
[00:40:09] Dean: Shared infrastructure occurs at multiple different levels. And it also varies historically around the world. So where I am in the UK, the four main mobile operators actually share in groups of two, a lot of their physical. Well, the actual physical towers, and I think some of the backhaul connections, and in Nordic markets, that's also pretty common and I know that the Chinese operators is doing it.
[00:40:31] So
[00:40:32] Matt: well, in the U S you've got companies like crown castle, that own lots of macro
[00:40:35] towers, but also have been buying a lot of fiber. And those are both shared
[00:40:38] Dean: Yeah. And so you've got shared this shared sort of passive infrastructure. So that's like the physical metal towers you've got and you've got perhaps the power center fiber back backhaul. In some cases there's a bunch of confusing acronyms that the mobile industry you can use is called a moron, which multiple operator radio network on multiple [00:41:00] operator.
[00:41:00] Core network. And there's a couple of others as well to talk about the different styles of infrastructure sharing, and they come with sort of different implications for security, different implications from a regulatory point of view in competition authorities. there's always a bit of a tension in most governments between competition authorities.
[00:41:18] On one hand, you'd love to see lots of independent physical infrastructure and the sort of. people who are trying to do the national broadband plan, who is saying, well, share if you can and get it to everyone, get broadband to everyone. and so what you end up with in some countries is, for example, again, you know, here in the UK, there's a new thing called the shared rural network, which essentially has been the government telling the forum, mobile carriers.
[00:41:43] If you improve your land area coverage of the UK with, in this case for GMO Fauji and you share towers, we will put this amount of money in to help you. so it's, this is like subs, it's a subsidized shared network, [00:42:00] but then in other markets, yeah, you've got, the tower companies who increasingly are, are doing fiber, and they're also doing increasingly they're buying our passage for in building coverage.
[00:42:08] In building is really interesting because it tends to be inherently shares, whether that is a distributed antenna system or whether it's, actually wifi for things like offload. the in-building environment becomes even more important 5g era because a lot of the flight G high-performance signals doesn't get through walls and windows.
[00:42:28] You need, and then the question, then you have this really the question of how do I deliver public services on private property who owns that work? Well, it depends on the type of type of building and it depends on the type of the country you're in and again, negotiating power. So I remember talking at an event a while back and they had.
[00:42:50] A hotel, CIO, hotel chain CIO in Asia. He said, you know, the funny thing is in, I think you said it was in Hong Kong. the network operators pay him and in [00:43:00] Thailand he has to pay the network operators and you couldn't work out. Why, why, why so different? But what well it's definitely happening is a whole new breed of what are called neutral host providers.
[00:43:10] A neutral host is, a shared network. A wholesale shared network. And there's also, you could argue the same in the, essentially the equivalent of a neutral, neutral co-location or data center company, but for a mobile network sometimes owning their own spectrum and allowing essentially the main carriers to roam onto them sometimes doing, yeah.
[00:43:32] Perhaps sort of and that, that gets,
[00:43:35] audio_only_16780290_Dean_Bubley: yeah.
[00:43:39] Dean: Yeah. Oh, the CBRS is really interesting. One. And the other option just to have, this sort of desegregated radio network, where you might have a shared antenna, but each of the operators brings its own virtualized baseband or control unit for the radio.
[00:43:57] And there's all sorts of interesting new models emerging. you know, [00:44:00] you might have utility companies doing your property companies doing it. there's local authorities and municipalities, and saying maybe we can get millimeter wave spectrum. And as a neutral host service down the high street, as a municipal, probably in conjunction with the tower company.
[00:44:17]one of the benefit here, incidentally, and this is me being a bit cynical is I do notice that, when a number of the carriers have spun off their tower companies and, or sold them to a tower company about a week later, they put a press release out saying we've hit our CO2 targets five years early.
[00:44:36] And I think it's because that's what as outsourcing their towers, they're outsourcing their carbon footprint as well.
[00:44:41] Matt: Yeah, that's interesting. so in the us, there seems to be an institutional resistance to shared infrastructure. you know, and it, I think that's eventually gonna sort of sort its way out when the dust settles in the, in accounting.
[00:44:55]but that may take some time. Are you seeing it different at different
[00:45:00] [00:44:59] Dean: Yeah. And I'll also a lot of it is to do with the fact that in the U S each carrier has got a really unusual, mixed bag of, spectrum. So they don't necessarily want to have their towers and radios in the same place necessarily because they've got different, a different sort of ran strategy. And someone says some of them own a lot of their own fiber as well.
[00:45:19] And population density is lower. So. It's easier to build, independent towers for each carrier in, in a lot of parts of the U S elsewhere in the world. There's a huge move to different forms of shared network, but especially because when five G while two things happen with five G first off the higher band spectrum means that it's almost impossible for every carrier to build out its own network.
[00:45:44] Even if it wanted to, particularly at millimeter wave.
[00:45:49] There's not enough fiber. There's not enough lampposts. Yeah. All this I'm particularly on . Well, I'm a potential client. I was talking to, I'm looking at doing this as a railway provider and the same, well, [00:46:00] you know, for millimeter wave, each of the carriers would have to have cell towers outside of the fence.
[00:46:06] So outside of the railway bland, and it probably wouldn't have the range. So actually what you do as a railway companies, you run your own neutral host network down the track side. Partly for your own use and partly you, you onboard, the, the MNOs as well, the mobile operators as tenants. so you've got all sorts of different sectors here.
[00:46:25] There's indoor, there's dense Metro areas, road and rail side, and also rural and rural is one where obviously a lot of countries are concerned about digital inclusion and network connectivity for, you know, rural communities and so on, but also agriculture and IOT. And there may not be enough people to justify any individual operator building out, a good network, but all the requirements, they might be able to have a shared one, either one of them shares and the others sort of tenants on the carrier number one's network or [00:47:00] to avoid the debate.
[00:47:01] They. Yeah. A third party sets up as a neutral provider, which could be a tower company. Or as I say, the municipality, there's a couple of the ones I know here that are almost like, sort of almost like a village is running their own cellular network, having interconnect agreements, and know there are a couple of companies and individuals who are really pushing that model.
[00:47:22] It could be for like an Island of Scotland, for example, whether they run their own network
[00:47:32] Matt: If, if you were a betting man, and I don't know if you are or not, where, where would you, where do you think this is going to net out?
[00:47:39] What's what models are gonna surface to the top.
[00:47:41] Dean: Everything is going to be a messy hybrid. I think the thing which wins. Yeah, everything is going to be deeply inelegant and based on sort of midterm [00:48:00] pragmatism, messed up by acquisitions. there's going to be awkwardness around local authorities. yeah, it's, it's going to be messy.
[00:48:09]maybe in some markets where you've got a Greenfield network, perhaps a dish in the U S or reliance geo in India, Rakuten in Japan, and then perhaps in China where you can, you know, they can mandate. Yeah. And particularly for things like the government does tell you, for example, building owners, you must have this enormous everywhere else.
[00:48:29] You got this real mess of overlapping jurisdictions and property rights and existing incumbency. And.
[00:48:37] Matt: wanting to get into it. I mean, what happens if Amazon buys dish?
[00:48:40] Dean: yeah. And then the other thing is, of course, is, is that that sort of democracy is that every four years someone gets in and wants to change everything halfway through. People want whatever the last competitor competitional authority was saying was acceptable or
[00:48:53] Matt: Right,
[00:48:54] Dean: in four years time, you can bet it's all gonna change.
[00:48:57] Matt: right. So you see high speed rail in the U S [00:49:00] yeah. So, so I want to switch topics a little. you kind of lit up a little when we mentioned CBRS, you said, Oh, see, breasts is really interesting. So can we, can you tell us a little about what makes it so interesting to
[00:49:12] Dean: Right. So I'm, I'm really interested in availability of localized spectrum, brilliant spectrum for, to build four G or five G networks. you know, historically most spectrum bands suitable for mobile have been awarded on either a national basis or in North America on fairly large areas. CBRS. Brings in a couple of different innovations.
[00:49:39] First is quite small sized, licensed chunks of spectrum for particular geographic areas. And we were talking, we're having this conversation just after the auction for that chunk of license spectrum kicked off. The other interesting thing is, is there's another. Tier of this band, which is sort of a bit [00:50:00] like wifi.
[00:50:00] It's not quite as unlicensed SIM permissionless, and everyone can use it, but it's open access. As long as, as you check with a database function called a SAS or a spectrum access system, essentially, you know, you. Yeah, avoid conflicts and yeah. And in particularly there's the incumbent networks, which is primarily the U S Navy, but also some other, some others.
[00:50:23]and you essentially, your small self says, can I use this bit of spectrum at this time at this power level? And that database says, yeah, go ahead. Actively. If you're in the middle of use or there's not too many aircraft carriers, so you should be good. Yeah, that's really interesting. Cause it w the potential users are five or six different categories.
[00:50:43] The existing mobile carriers might use it to supplement their existing spectrum in high density locations. Fixed wireless access. wireless service providers might use it particularly for rural coverage, to build out a fixed 4g or five G network. The cable [00:51:00] companies are interested in it to build wireless, almost like the extensions to their cable and fiber plants.
[00:51:06] And then enterprises are interested in it because let's say I've got a massive oil field somewhere and I've got, you know, 500 oil Wells and there's frankly, no one lives there. This means I can have my own private wireless wireless network in protected spectrum, collecting data and doing all my IOT and environment protection.
[00:51:26] Run on my private cellular network. without having to go to the carrier and pay per month or per gigabyte. and there's a bunch of other use cases for municipality's. yeah, this neutral host and indoor model and sort of shared access license. I was saying maybe you're an airport authority. You want to run a private cellular network where the airlines and the oil company and is, and the catering companies and the maintenance companies.
[00:51:49] They're yours. Tenants and you act as a, a little local airport service provider or a portal, for instance. So it would be really interesting to see who ends up with both [00:52:00] the licensed bits of this ban, but also the, this unlicensed section. And to me, this is one of the leading examples around the world of this local wireless, spectrum, program, the UK, Germany, Japan, France, and a number of other countries also have something vaguely similar.
[00:52:22] So if you are an airport authority in lots of countries, now you can build your own private 4g or funky network. Some of the bands and some of the rule sets of more in tune to, to this fixed wireless access model. Some are more aimed at, say in Germany for industrial companies wanting to have a private five G network, to run their robots and industrial automation systems in the factory.
[00:52:45]France tends to be more in that replace things are all older sort of two way radio systems like Tetra for critical communications that ports and airports, but there's a lot of innovation going on. And so we're going from a world [00:53:00] where there were now today. There's probably about 800 mobile carriers around the world.
[00:53:04] We had this conversation in five years' time. We, there may well be 8,000 or even 80,000.
[00:53:09] Matt: eight.
[00:53:12] Dean: what you probably have is you'll probably have eight big ones and 8,000 small ones.
[00:53:16] Matt: Right. Interesting. The long tail will be really long.
[00:53:19] Dean: Yeah. I saw an example demo. someone did where they got a local spectrum license from off come here, here in the UK.
[00:53:26]set up for an office in the center of, of London. took them two weeks to get a spectrum license for, I think it was like 10 megahertz spectrum, which is enough to run a basic 4g network and it costs them essentially a hundred dollars for the spectrum license. and they had a small cell. A little tiny core network running on a server, the size of a shoe box.
[00:53:47] And they were demonstrating using fresh out of the box, Android phones with their own network name on a licensed local license network to make phone calls. Now that's not [00:54:00] the world's most amazing thing, but the speed that it took them two weeks to get the license and a hundred bucks.
[00:54:06] Matt: Yeah, it's almost like the cellular network could, you know, in some future state be like infrastructure as a service.
[00:54:13] Dean: yeah. Or frankly, if the aim is the, some people I speak to to make it easy, to set up as a wifi network. Now, obviously it won't. If I know at one level you can just go to the store and buy a, a wifi access point or an enterprise. It's a bit more of a sophisticated, you have a professional installation, but if you can, and in fact, a lot of the wifi.
[00:54:32] Integration and installation companies, aren't trying to do CBRS accreditation, because what that means is they could come up to your office building and potentially put in a private cellular network, as well as a private wifi network. and that then you could use that for different sets of use cases.
[00:54:48] So, yeah, wifi obviously gets congested is very good for certain things, less good for others. But imagine you have 'em on a retailer. you want to offer wifi or an airport. You want to offer [00:55:00] wifi to your customers and maybe feel, I don't know, display panels, but you would really like to have your payment terminals you're running on a private cellular network because the less prone to interference.
[00:55:15] Exactly. Exactly. So there's various reasons why I want to do both. and so what I see is a future world where. Particularly indoors, you will have indoor wifi for laptops and for screens and for voice assistance and all of the devices, which are wifi powers, you will want to have public mobile networks working indoors,
[00:55:39] Matt: a building. Sure.
[00:55:40] Dean: visitors phone.
[00:55:41] So you don't know what network.
[00:55:42] Matt: Yes.
[00:55:44] Dean: you might want to have a private cellular network.
[00:55:48] audio_only_16780290_Dean_Bubley: Yeah.
[00:55:50] Dean: all sorts of things. Are you, maybe it's a no. And you want to run your security cameras and yeah. Gaming machines, on that. So more, perhaps more critical things. It might have a better [00:56:00] range, but, yeah.
[00:56:01] And then the interesting question is where edge compute fits in this and particularly in industrial. Environments. There's an argument in that, say a factory. So have, your robots and your industrial automation systems and your quality inspection cameras connected to your private cellular network. And then the question is where do you do the, I don't know the video analytics for the welding machine to work out.
[00:56:25] If a weld is good or bad, do you run that on a server in your factory or in the parking lot? Or nearby, you know, a few minutes, seconds up the road,
[00:56:38] Matt: Yeah. Well, in fact, you know, it's interesting. One of the, what are the emerging use cases that I'm seeing in our business is what we've started to call near prem.
[00:56:46] And this is this idea that if you're building all these data centers and networks, you know, out in the field, You could draw a 50 kilometer ring around every data center and identify factories and whatnot. And so you look at a factory saying, well, [00:57:00] I need discrete latency because I'm driving a robotic arm or something, but you know, but maybe it's three milliseconds and not, you know, 60 microseconds.
[00:57:08]and I can actually, I don't have to drop the data center in my parking lot. I can put some servers in a co-location facility and have a direct fiber run.
[00:57:21] Dean: Yeah. Yeah. And I think this, this sweet sort of sort of 5,000 kilometers, something like that. And that interestingly, that also works with some carriers and countries, you know, say in Europe where the size of the country means that you can essentially cover a country with. 10 or 20 data centers that starts to become more of a, an opportunity for service right now.
[00:57:43] Obviously there's some questions there, which is if I've got a, a multibillion dollar factory, I don't want to run the risk of a fiber cut taking out my near edge. A computer, but if I'm a small, a smaller facility, then maybe I can have a fiber and I have a five G fixed wireless [00:58:00] as backup or something like that as a fail over.
[00:58:03] And so I think that it's going to depend on the application, but I can certainly see that near prime makes a lot of sense. It also makes sense. If you are, let's say a retailer
[00:58:16] Matt: Yeah. Why would you put a, put a server farm in each store? Well, and the other is other two other use cases that I've, that I've seen, which is so hospitals and, venues. so. You don't want to, when a, when a venue opens up for the public, the ability to act access server equipment on site goes way down.
[00:58:32] It becomes a security issue. whether it's a, a motor Raceway or a baseball stadium and hospitals, and you can see with COVID-19 like you can't get in there often cases. And so having it, you know, quote unquote, offsite, but nearby, or having an alternative outside in the Bay, you know, provides potentially some access that you wouldn't normally be able to have in emergency
[00:58:51] Dean: Yeah, that certainly makes sense. I mean, there's, there's a lot of sensitive sites now. It depends the more, the most sensitive we'll probably still want on prem if it's out of the military [00:59:00] or something like that, but there's probably other ones which are, it could be small cities. And particularly with, as you say, the pandemic and what happens in the future, if there's another way you don't want to have to have your.
[00:59:12] Compute resource off limits or needing certification or a piece of paper pay paperwork from someone who's working from home and who isn't there. Yeah. I mean, this, this is, I actually talked to a couple of data center companies recently who were saying that the problem they have at the moment is there's lots of demand for new data center capacity, whether that's hyper scale or small scale.
[00:59:34]but they need various bits of paper to go and have a permit to. Yeah, send an excavator out to dig up the ground to install. but the town hall is shot and everyone's working from home and actually there's all these sort of administrative bottlenecks on it. So yeah, the more you can do things off prem automate them, then there's going to be some glitches.
[00:59:54] I think that makes a lot of sense. And I like this idea of near prem because, and that's [01:00:00] also potentially a good meeting point or meet me point for all the different carriers as well. And this is what it was. It was a post. Yeah, there's a, there's a question then, is the assumption in the mobile industry, is that the edge goes in the net?
[01:00:17]as we were saying before, whether it's for, offload or anything else, or a magnet. I actually think that possibly the network goes in the edge and actually what you might well find is these localized data centers that have chunks of each of many different network operators, infrastructure in a, a neutral Cola.
[01:00:37] And so it's almost like terms of paradigm around, it's not necessarily what. I think people in the telecoms industry would like to hear, because although it's less, these days, a couple of years ago, in the mobile industry saying, Oh, we can create a distributed quasi AWS with their own edge.
[01:00:57] Yeah. Yeah. I wish I was unconvinced by, [01:01:00] but I, I think this is interesting question here is for a lot of my clients are in telecoms and I think that they will end up being both sellers and buyers. Yeah. And depending what it's for where they are. Yeah. Whether it's for the bits of the network, which really need to have them owning it, or whether it's for running them billing system. And actually I can quite imagine companies running their customer care software on public cloud. and so I think that the interesting, this goes across board for lots of other things, is that the Telecomes industry has to wake up to this idea that actually there's a marketplace.
[01:01:42] And sometimes it makes sense to be on the buying side and sometimes on the selling side,
[01:01:47] Matt: That is a really insightful way of looking at it. Yeah. So, so, so one of the things, things that I wanted to ask you back, cause again, I'm sure you have an opinion about it and it's, you know, if, if you look over the last [01:02:00] decade, the early part of last decade, the telcos have been licking their wounds over the OTT providers and the cloud providers in particular, like we're just dumb bit pipes.
[01:02:09] And when we roll out five G no way, are we going to be done bit pipes again, and that created a lot of tension between the cloud providers and the telcos, but. at least in the us, we've reached, you know, when the last 12 months seen a lot of partnerships between the cloud providers and the telcos,
[01:02:25] audio_only_16780290_Dean_Bubley: Well,
[01:02:25]Dean: when I look at say, I mean, so you're thinking of things like, imagine like Amazon wavelength, earliest wavelength in their partnership with Verizon, Verizon, and I think Volvo as. That's. I mean, that makes sense until you suddenly, I saw a presentation recently on AWS is edge strategy and they've put up a chart on one of the events and they have like 10 different edge products and services of which wavelength is one of them.
[01:02:52] And so essentially they're covering all the bases. And so. Yeah, there will be some developers for some locations in some [01:03:00] places that want to have something inside a five G network, but there's others who will want to have an on prem device. There's some, they've got some real time operating system that you can do in a Silicon level, right up to the, sort of, just standard availability zones or increasing the.
[01:03:17] The reach on. And so I think that those hyperscale companies, they are doing everything well, if you actually look at the portfolio, they're looking at going, yeah,
[01:03:35] Matt: Yeah, that's really interesting. Well, and I also think the telcos have realized that, while they may have their own. Let's call it a cloud. So the ability for third parties to deploy workloads, that part of what they need to do to be successful is to deliver the, all the applications that their end users want.
[01:03:56] And most you can't ignore the 10 million [01:04:00] developers that want to write for the major
[01:04:01] clouds. And aren't going to write for
[01:04:05] Dean: Yeah, absolutely. Yeah. I, I don't think this is, this is the other thing, is that yeah, if you're looking at, I dunno, an industrial company, they've got a massive sort of it fleet of applications and services and some of their individual workloads in some of their applications will need to work at the edge, but others will be.
[01:04:24] Analytics storage AI, it's the R and D function that these, the storage three dimensional models. Yeah. That perhaps it doesn't need to be age based and they're not going to sort of. Orient their entire it development area, environment around the handful of workloads that happened to need this capability.
[01:04:44] You know, they're, they're, they're, they're the adjuncts that there might be important, but you're not gonna use that as the it's like the tail that wags the dog. Yeah. It's a, it's a, yeah. And I think that bits of the tech industry I'm speaking to do get that, and I think they are seeing those [01:05:00] partnerships.
[01:05:00] And I think that what you end up with is.
[01:05:09] Matt: Yeah, I think they have to. I mean, you can't ignore all the apps that are on Azure. Well, just because you have a partnership with Amazon. Well, and also this brings us back to, you know, our, know, imagine future state where you have these neutral, facilities, and you've got one rack with the telcos one's equipment, another rack with telco equipment and another rack with Amazon's equipment and other rec with Azure equipment.
[01:05:30] And you've got this, you know, software-based interconnect that is allowing them to peer or, or exchange data, whether it's a paid relationship or a settlement free relationship.
[01:05:44] Dean: Yes. And in fact, you not just them, you may also find those enterprises and you might find that it's a, a large car manufacturer. Who's saying, well, actually we want to have our own connectivity or a utility company, or, yeah. So there's actually a. [01:06:00] Yeah, I, yeah. All particularly some of these, these companies who are getting CBRS licenses, they're telcos too.
[01:06:05] Yeah. And I think there's this, this idea of sort of legacy spectrum privilege, if you like, where there's this assumption
[01:06:12] that just because you've been running mobile networks in perpetuity,
[01:06:16] you are entitled to all of the connectivity and the cloud functions
[01:06:21] Matt: Well, we, we may, we may end up in a, in a world where, where if you know, CBRS shared spectrum works at scale, which it looks like it will, based on the test, that we may wish that all the spectrum was treated like that. And I don't know, we might, it might eminent domain it back with the pay everybody, but the government might repurchase the spectrum and make it shared
[01:06:43] Dean: I mean, I would be the in fact that doing something like that, which is there's some approach, which is it's a bit like use it or lose it.
[01:06:54] Matt: Yes, right. Like water rights in the U
[01:06:56] Dean: or use it or use it or let someone, and I was borrowing [01:07:00] in the meantime, which is, you know, if you are running a wind farm on a remote Hilltop in Wales and you want to connect your wind turbines and none of the carriers, although they in theory have national licenses for business spectrum, none of them have built out coverage.
[01:07:14] There. You can actually go to the regulators saying. Can I have it. and there's actually now a process by which that secondary authorization, which is another form of the infrastructure sharing in a way Canada can operate. So I think you'll, you'll see that as a sort of an interim stage. I'm not sure that we're going to end up with this sort of, you know, libertarian ideal of everything becoming Speco spectrum as a service.
[01:07:37] Matt: telcos will have an opinion on this.
[01:07:39] audio_only_16780290_Dean_Bubley: But
[01:07:39] Dean: what I do see that a number of companies that are moving towards that way of having spectrum as a service and perhaps dynamic spectrum databases and online allocations. Yeah. I don't think, I don't think we need like, options markets on this.
[01:07:53] Matt: reallocated spectrum in the U S you look at when the migration to HDTV. I
[01:07:57] Dean: yeah. With C-band coming up, which is [01:08:00] where essentially the government has gone to all the sudden, initially the satellite industry came to the FCC saying, Hey, we reckon we can rejuggle our spectrum use.
[01:08:08] And then. You sell the rest of it and give us $10 billion. And the FCC says, good idea, but we're not paying you that much. I paraphrase, but that's roughly how it, how, how, how it's going. And so I think the satellite industry got ahead of the curve on that, realizing that it can almost unilaterally decide to sort of reorganize.
[01:08:27] The the military, and the government and a number of countries are doing that. and there's various ways that different, administrations can incentivize this by essentially making government departments pay for the spectrum they use,
[01:08:43] Matt: Yeah. So I have one last question, Dean, and then we can wrap. so something you mentioned earlier that I think very few people, will pick up on, but I'd like to double click on it. You mentioned, MEC. And it started as mobile edge computing. Now it's multi-access edge computing, which is an interesting shift.
[01:08:59] So [01:09:00] what do they mean by multi access and how does 5g relate to wireline?
[01:09:05] technology and we almost always talk about it as a wireless technology. How, how did that, how does that all fit
[01:09:10] Dean: Well, a lot of it comes down to a lot of the carriers around the world are mobile and fixed providers and they are trying to create. Irrespective of edge computing, they're trying to create converged services. So let's say I am Telefonica in Spain or, yeah. And I want to provide a mobile video streaming service, both to my home users on home broadband and to their mobile devices.
[01:09:34] And maybe there's a bunch of other features around smart home or security cameras, or I do a deal for. Yeah, I don't know, video doorbells or whatever it is. and so then I want to have some formal compute platform, firstly, which is likely to be physically located in my local or city level. sometimes Metro service knows which essentially is like, it's like a [01:10:00] central office plus a mobile aggregation physical one physical sites.
[01:10:04] And so then that point, the thing, well, we'd love to have compute in that. City-level facility, not just for the mobile traffic being aggregated, but also for the fixed. And while we're at it, can we use the same core network architecture for fixed and mobile? So I think that that sort of, that's a Mo that trend towards, service convergence, bigger trend than edge computing.
[01:10:31] And to some extent that comes first because it's about the access network is the people, you know, if you, if you look at where the CapEx goes in a fixed or a mobile network, or certainly a converge one, it's the radio network and it's digging up the road to put fiber. So those groups to some extent, drive the architecture for everything else.
[01:10:49] And if they're putting all of their stuff in, in a multi-access. City edge node. That's a sensible place to put your edge compute as well. And also to be honest, [01:11:00] a question I've asked for, for ages is edge computing is going to be such a big deal, don't we see fixed edge computing already?
[01:11:12] There's Tom, we are now, but yeah, but no, one's been talking about fixed edge computing for the last 10 years. Yeah. Yeah, yeah. But yeah, but, but also the near the near edge, yeah, if you look at all your, what are the things, you know, sort of toward the industrial companies, the most enterprises have one or two fibers into the building and you would have thought if, if, if this millisecond latency thing was such a big deal,
[01:11:41] Matt: Well, I think the catalyst of the cloud providers making it easy to run workloads in those locations could radically transform it. But it's a, it's a, provocate, it's a provocative question. It's a provocative question. 18, you know, this has been one of the most enjoyable conversations I've had so far in this podcast.
[01:11:56] I really appreciate you coming on. I would definitely like to have you on [01:12:00] again in a little bit, to like catch up and chat about stuff. this has been terrific.
[01:12:07]Dean: the easiest way is on either Twitter or LinkedIn Twitter. I'm disruptive Dean. on LinkedIn, I'm usually talking about various things, but you can add me and if you're in the industry, I'll connect with you and, via that route also, my website is DeanBubley.com, but I don't update it very often, but yeah,
[01:12:25] Twitter and LinkedIn, probably the easiest.
[01:12:29] Thanks.
[01:12:29]