Today’s episode features an interview between Matt Trifiro and Alex Reznik, Distinguished Technologist at Hewlett Packard Enterprise and Chair of ETSI’s MEC ISG. In this interview, Alex explains the role of standards bodies like ETSI, how standards drive market adoption, and why they are important for the future of the edge intersection between telcos, cloud providers, and developers. The Edge Computing World Conference is October 12-15th, 2020. Fully virtual. Register at www.edgecomputingworld.com and use the promo code OVERTHEEDGE for 30% off the Edge Executive Conference
Today’s episode features an interview between Matt Trifiro and Alex Reznik, Distinguished Technologist at Hewlett Packard Enterprise, and Chair of ETSI’s MEC ISG.
Alex literally wrote the book on Multi-Access Edge Computing, and as Chair of ETSI MEC he is helping to pioneer the industry standards for enabling the Intelligent Edge.
In this interview, Alex explains the role of standards bodies like ETSI, how standards drive market adoption, and why they are important for the future of the edge intersection between telcos, cloud providers, and developers.
Key Quotes
“In the final state of this MEC public edge computing game, there [will be] a presence from most of the major cloud providers in most of the operator networks.”
“If telcos are going to succeed in capturing value, they will have standards in a space where ETSI MEC plays. It doesn't mean they have to adopt ETSI MEC standards, but they will have to agree on a set of standards in that space. Our value is, we're the only standard right now that plays in that space. So you can either go and form a new group and start from scratch, or you can go with ETSI MEC.”
“With 5G, specifically because so many of the 5G applications need the edge, and edge means location, [the telcos] have leverage. Amazon and Microsoft and lots of other people have to talk directly to the telcos. That's real leverage. Now what are they going to do with it?”
“Show the operators the money. They want to see the money. They don't want to see a strategy talk about how much money the hyperscalers are making on the cloud and what portion of that market they can address and what the TAM for edge computing is...they want to see the money in six months.”
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 Seagate Technology. Seagate’s new CORTX Intelligent Object Storage Software is 100% open source. It enables efficient capture and consolidation of massive, unstructured data sets for the lowest cost per petabyte. Learn more and join the community at seagate.com
Links
Alex's book, "Multi-Access Edge Computing In Action"
CLICK HERE to Register for the Edge Computing World Conference, October 12-15th, 2020. Fully virtual. Use the promo code OVERTHEEDGE for 30% off the Edge Executive Conference
[00:00:00] Matt: Hi, this is Matt Trifiro, CMO of edge infrastructure company, vapor IO, and co-chair of the Linux foundation's state of the edge project. Today. I'm here with Alex Reznik, distinguished technologist at Hewlett Packard enterprise. he also chairs ETSI, MEC industry specification group, we're going to talk to Alex about his career and technology, his current roles at HPE and everything. Edge computing. Hi Alex, how are you doing today? So you're one of the edge computing. Oh, gee. you've been doing it for awhile. but before we really get an edge computing, I'm super curious. Like how did you even get into technology? Cause you have an engineering background, right?
[00:00:37] Alex: Yeah. yeah, I was good in math when I was going to college and I sucked at dealing with blood, and the. Next best career was to go into he cause I was in New York city, but the school I went to didn't have a computer science major. I [00:01:00] wound up doing ye and at some point, the chair of the department taught a required course introduction to communications theory and.
[00:01:09] Matt: John.
[00:01:12] Alex: No. Yeah. Now his name is Mel Sandler, but, he spent a lot of time doing something for the CAA, with the CIA. He was never too specific, but his course was real fun. And so the whole Lakey Curry or a plan got sidetracked and doing comms. and that's how I wound up in this industry.
[00:01:33] Matt: Oh, that's really interesting. did you ever do any telecommunication engineering or
[00:01:38] have
[00:01:38]Alex: no, I did. I did. Yeah. So the way I wound up in edge computing is through the M part of it, which really comes out of the telecom. I, my whole career was, to some extent, still is the telecom side and it's just too well with the exception of the first job, which really was it.
[00:01:58]and then, now [00:02:00] there's convergence of the two, which just makes it that much more fun.
[00:02:03] Matt: That's awesome. So how does, for our listeners who don't maybe understand the whole world of telecom, what is ETSI?
[00:02:11]Alex: ETSI is officially the European telecommunication standards Institute. is what it stands for. And it does two things. one is it serves as an almost regulatory bodies. So they publish things. They call the harmonize standards and those are the standards or other policies that harmonize things like spectrum policy, et cetera, across the European States, something that's.
[00:02:39] Uniquely necessary in Europe because there's this patchwork of fairly small countries that are highly integrated, but yet really fully sovereign. So that's, and the other thing that they've gone is technology standards and they used to do with just themselves and just for Europe. So GSM, right? The fame is GSM the [00:03:00] first mobile standard.
[00:03:02] Matt: To be confused with the GSM a.
[00:03:05] Alex: Yes. Not to be confused with a GMA, although that's where the name comes from. But a GSM wasn't that substandard initially, what has happened with ETSI is with the formation of squeegee PP, which ETSI drove. And in fact still does the logistics for three GPP, They've lost a lot of that.
[00:03:25] Or there are other transition that leadership and the mobile technology standards over the three GPP, That had to happen for it to become a global standard, to get the buy in of the other seven. Global organizations. Yeah. What it also did is it diminished BETSI for a long time as an international standard setting body.
[00:03:48] And then over the last half a decade, within a fee, they came roaring
[00:03:53] back. And so
[00:03:54] for all intensive purposes, right? C is now an [00:04:00] international technical, the standard setting body really, truly global. In addition to really being a European kind of regulatory state standards, body, and I'd probably state there.
[00:04:13] You're being the regulatory role somewhat indirectly. Cause I don't know exactly. How much of what they do has advisory and how much of it has, sorry, power. But,EC so don't quote me on that, but they have that role, which is European and technology standards role, which is where ETSI, MEC
[00:04:30] fits really
[00:04:32] global.
[00:04:32]Matt: how does ETSI relate to the five G standards?
[00:04:37] Alex: So the official five G mobile technology is being standardized by three GPP. Of course, ETSI is structurally three. GPP is a consortium of, I believe, seven international or geographical standards organizations, which are at sea at, [00:05:00] A Korean one, a Chinese one. I forget all the names. So it formally relates to it that way.
[00:05:04]But when people talk about five G standards, they very often talk about things beyond what's three GPP specifies, which is the mobile network. And so in some sense, the ice standards that we develop and MC okay. Our five G standards, the ETSI on the standards, They're really necessary to make five G over reality
[00:05:25] Matt: I certainly think of them as part of the five G standard, even if they're not in the same
[00:05:28]Alex: And at sea has been producing them. So we're I'd say we're part of the mosaic of standardization. That's necessary to make five G or reality. And when I say we, ETSI,
[00:05:40] Matt: Yeah, or well, and everybody else too, we're all working towards making that a standard,
[00:05:45]Alex: From a standards. There's the standards mosaic. then there's the people that actually built stuff. And I don't mean that facetiously right. Standards have a role, but their role is not to build stuff.
[00:05:56] Matt: Yeah. Yeah, that's true. so your current [00:06:00] role with ETSI is the chair of the MEC ISG, the international standards group. and that stands for multi-access edge computing. Originally is mobile edge computing. We're going to get into that, but let's, let's cover what the role is today. So how long have you been the chair and what is your role as the chair of
[00:06:17] Alex: So I've been the chair. It's now about spina half years. the group was. Close to six years old nerd Sprecher was the founding chair. And then after two years she moved on. and so I took over that position. So it's been awhile, it's been a three and a half years.
[00:06:36] Matt: Yeah. and a, what is your role? what do you do? cause you, have
[00:06:38] you
[00:06:39] worked for Hewlett Packard and Hewlett Packard
[00:06:41] Alex: What does a chair do? And the standard is a go. So I'll give you my, My view of what they're all of a standards, group chairs. It is specific perfectly to not do on your technical work. That's what the people that count to the standards go for. So you're [00:07:00] a facilitator, right? Your job is to make sure that the group run smoothly.
[00:07:06] And so that during the technical discussions, which can get. fairly contentious at times because you've got different companies bringing in. Ideas, but they're not just a year. So there's a significant investment behind their own way of doing things. You have to facilitate an atmosphere in which they can arrive at a mutually acceptable compromise, right?
[00:07:31] That's the goal mutual, make anyone happy, choose to, or arrive at something that's
[00:07:37] mutually acceptable
[00:07:38]Matt: it's a little bit like Congress.
[00:07:40]Alex: the, not nowadays,
[00:07:42] Matt: Yeah, it's true, right? Congress is
[00:07:44] Alex: that should be their goal. But it's not. that's one goal and the other role is you're the promoter and chief for that particular standards party.
[00:07:54] So, that's
[00:07:54] what I try
[00:07:55] and
[00:07:55] do
[00:07:55]
[00:07:55] Matt: some of us call that marketing and sales. How does that feel like as an
[00:07:59]Alex: yes. [00:08:00]
[00:08:00]the interesting thing is right for the last four and a half years, I've been at HB and it's a technical sales position that I've been at and not an engineering position. So the two kind of jive fairly
[00:08:13] Matt: Yeah, I can see how much of your time does it take up a percentage?
[00:08:20] Alex: Yeah, I enjoy doing it. I really don't keep track of it. And the other part of it, it is, of course, if you're doing sales and business development, It's not like it's a nine to five, so you go, Oh, every day, three hours of it gets spent on ETSI MEC it's. So I don't know, I don't really know how much of my time is taken up on my primary.
[00:08:40]job, I call it an HPE sponsored hobby. They certainly enabled me to do it, but,
[00:08:47]I do know the chairmanship, the job is man. The job was very difficult and it's worth every single penny that HPE is paying me
[00:08:57]Matt: So they're getting more value than they expected.
[00:08:59] Alex: are [00:09:00] they're getting more value. Absolutely.
[00:09:01]Matt: that's really excellent. Especially in this day. what is multi-access edge computing?
[00:09:06]Alex: So shameless self promotion, I wrote a book, well i co wrote with a couple of other
[00:09:12] Matt: Available on Amazon.
[00:09:14] We'll put it in the show
[00:09:14] Alex: if you read, the first chapter, which I guess you'll not find out, I wrote that one for better or worse. I talk about it this way. I go look at it from the edge, the C stands for computing or cloud, right?
[00:09:29] So it's first and foremost cloud, then you get to the E and it's the edge. So it's an edge cloud and the am, whatever you choose to attach to, it really indicates that you talk, that you're taking this edge cloud and it's located at somewhere in a communication service providers network. So it's not an edge cloud in an enterprises on-prem location.
[00:09:57] It is somehow a point part [00:10:00] of what. Before five G and we still call a service provider in that work. I would argue it's time to start calling it a teleco cloud. And so that's what MEC is. So I always tell people, remember that first and foremost, it's edge computing. It's not something different than the edge computing.
[00:10:19] It is. Edge computing is just. You're going to do some things differently, or you're going to do some more things with it because of where it's located and the fact that enables you to do some other things.
[00:10:33] Matt: Yeah, it's interesting. You call it a telco cloud, because. the shift in the name in 2017, I think when it was originally called mobile edge computing, which makes sense, it was primarily for mobile phones. That's primarily where ETSI does its work. And now it's, multi-access edge computing, which implies wireline and other access network methods.
[00:10:54]and so why do you call it a telco cloud as opposed to
[00:11:00] [00:10:59] Alex: communication server was provided the club. because, so
[00:11:07] I'm going to pick an Amazon cause I actually, I want to Pick an Amazon, right? aside wavelengths what's edge to Amazon. It's an outpost. So an outpost can't just be deployed in a telco cloud is it was designed for an enterprise to have on prem. But when you look at the whole management of it rights, you can the gear into a telco cloud, but unfortunately you can't position it as something public.
[00:11:39] Telcos edge cloud customers can, he can use cause it's a very different way of management. what an outposts was really designed to do is if I'm running my own smaller midsize company and I take advantage of AWS and I need something on prem, but I want to feel like I actually have an AWS on-prem right.
[00:11:58] I want that [00:12:00] seamless management. I don't. I want to not worry about having to target a different cloud for my apps, for the edge, et cetera. That's what an outpost was designed to me. same thing with Microsoft, same thing with Google, right? That's an edge cloud. And if you want, then you ask them what is edge.
[00:12:17] They would say, this is edge a, when you look at the local zones that AWS is deploying, that's still that kind of an edge. It's not a telco edge. It's not an. Edge that is, intimately part of a telecom network is designed to specifically be public right. And connected to directly Seville the connectivity that a teleco customer provides wireless a wire describing them irrelevant.
[00:12:44]So that's where the thing you need to keep in mind where the location of the edges
[00:12:51] and what you can do with it.
[00:12:52] Matt: Yeah. So you mentioned wavelength and you dismissed it, but isn't wavelength like
[00:12:59] Alex: So [00:13:00] wavelengths is designed for MEC. Right now it's Amazon's And yes, it's build an outpost, but I think what people will find is the way you interact with wavelength is going to be very different than the way you would interact with an outpost on your own premises. Even though most of the gear I'm the needs, at least in the beginning is going to be the same.
[00:13:22]so it's a different
[00:13:24] Matt: Yeah. Yeah. But I understand the outputs form factor makes it easy for them to,
[00:13:29] well, on the PowerPoint, it makes it easier. It's just a, an outpost box.
[00:13:32] Alex: well, it's certainly easier than wheeling in their whole
[00:13:35] Matt: Yeah. you've clearly been in the technology industry long enough to
[00:13:40] Alex: is yes. In that particular case, that is far from an optimal form factor, in my opinion for MEC. But again, I'm sure there's plenty of faults in it, especially at Amazon who will disagree with me.
[00:13:54] Matt: Sure. Sure. and it doesn't necessarily have to be the longterm plan either. If it turns out that's something that's in high demand, they [00:14:00] will, they're very customer focused and they'll deliver something like that. And there are other cloud providers that are also
[00:14:04]Alex: And to be, to give Amazon credit. Yes. when they see demand and they see that they did something, not quite right, they fix it very
[00:14:13] Matt: Yeah. I'll play a game here, which is, I'm going to tell you what I think the origin story of MEC is. and then I want you to correct me cause I'm sure I'm wrong. was told that the origins of MEC actually came from Nokia. Nokia had a program, they called liquid apps.
[00:14:27] And the idea behind liquid apps is, again, this, for our readers, not listener, not for you. but, Telco networks historically have been built out of dedicated appliances. And these were boxes that you'd buy from Nokia or back of the day, Alcatel, Lucent, or Erickson and all of that.
[00:14:43] And, you would drop it in and it like that was it. But you look inside these things, they tend to have Intel servers and Nokia recognized a potential opportunity. that if they over-provisioned, the capabilities and these boxes that potentially could run applications on, it could run other people's workloads.
[00:15:00] [00:14:59] And if you stitched a bunch of these together, you might actually have, and somehow, and then magic happened, that became part of the, that became the origin of ETSI MEC. But I don't know if that's true. do you have a version of the origin story that aligns with that, or is maybe different?
[00:15:14] Alex: So clearly Nokia's liquid apps program was. A significant catalyst, because if you go back, there is a formal origin. There was a white paper that was published by ETSI. bite people formally through ETSI, which then through all of the establishment of the group. When, if you look at the ad who published the white paper, it was no care.
[00:15:40] And I know those folks, it was out of the group that did liquid apps as well as wildly, as well as Vodafone. What you're saying is now I know Nokia, so I don't know if it's true, but it sounds very plausible as part [00:16:00] of the origin story. You have to remember. The other thing that has happened by then is NFV has started gaining traction.
[00:16:08] And so
[00:16:09] Matt: at? The white paper was published.
[00:16:10] Alex: was 2013,
[00:16:12] Matt: definitely.
[00:16:14]Alex: actually I've maybe no. 20,
[00:16:17] 2014,
[00:16:18] Matt: Okay.
[00:16:18] Alex: I think so.
[00:16:20]you have these apps running, they're still telco apps. You still can't manage them the way OpenStack manages a cloud, you need more. Which is why NFE is there. So you needed to have that in place as a standard in order to drive max.
[00:16:35]It's presence enabled a Nokia to go public with it. But the other thing that happened is again, Nokia found likeminded companies that said, yes, this is a direction that now makes sense to explore, not just among the operators, but also among the vendors.
[00:16:53]Matt: Yeah, and it does make sense. you really have a vision of attracting developers that want to deploy something globally, [00:17:00] it's going to have to be, have some level of vendor neutrality and some level of network neutrality. And so a standards body is
[00:17:07] probably one of the best ways to accomplish that.
[00:17:10]Alex: and yes, Nokia, you could say Nokia was definitely a pioneer in this way of thinking. but if they were the only one or let's put this, I don't know how far back liquid apps goes for it. may have some inkling. But when ETSI mech happened, even if Nokia was a pioneer, at some point, they were no longer the only ones that we're thinking about this direction and liquid apps were public by that point.
[00:17:36] Anyway, they've made of noise about it
[00:17:42] anyway, so people had known, right? so this kind of thinking for the edge had caught out by that.
[00:17:46] Matt: Yeah. and I, and your point about NFA makes a lot of sense because, when I've gone back and looked at some of the original liquid apps, materials, which are hard to find, but sometimes I found some YouTube videos and stuff, again, they're still in the early days in this mindset of [00:18:00] it's running onside inside our appliance, that's deployed in NFV completely.
[00:18:05] Appended the idea of an appliance. It's no. We're going to have sure. We're going to have, specialized RF, circuitry and stuff, but ultimately a bunch of these network functions are going to run on white box servers in a facility that looks. Something like a data center. and presumably has the kind of cloud orchestration capabilities that an OpenStack or some other solution would provide.
[00:18:25] And so that seems to me to make a lot of sense. And so let's tease that out a little bit. so there's a lot of functionality in let's over-simplify, let's call it the Ram the radio access network, which you know, is combination of the base band unit. The radio had a bunch of other things, that the connection between the computer that's running.
[00:18:46] The MEC application and computer or hardware that's running the network function for the base band unit, for whatever else for for all these other things, they [00:19:00] don't have to be the same machine. And increasingly they're probably not going to be the main machine. is that true? Say machine not the main machine. Yeah. and so that makes sense, because then you can isolate the next order functions. You can, they have discreet timing that, VMs could throw off and things like that. and again, as a, an occasional student of the evolution of MEC, I see, I think I noticed a point where that transition became.
[00:19:25]really profound in that, ETSI standardized around an API, as opposed to saying here's like the cloud it's gonna run on. It's no. Here's how you can have cloud workloads running in the same facility on different computers or even in nearby adjacent facilities that are, a sub one millisecond fiber run apart.
[00:19:46]but are talking to the radio network over an API. Am I accurate and observed in that observation.
[00:19:53]Alex: yeah. so for a while, there was an expectation that. We're going to standardize [00:20:00] something functionally on something for the telco cloud. And I think that's driven out of looking at say what's three GPP standardizes. Which is a lot of functionality. but really, That was a mistaken impression for two reasons.
[00:20:17] One is just, what is the role of standards? The role of standards is interoperability, right? And. When you're talking about devices that are communicating over a physical media, there has to be some level of real functionality definition to ensure that interoperability, as you start going up and you're running on what is standard computer architectures and standard networking architectures, right?
[00:20:48] Whether it's a Sianna, whatever that. Need for functional standardization goes away. and I started being involved with standards many years ago, an [00:21:00] old I Tripoli eight Oh two 11 standards. Hadn't told me that standards that forget that their only reason for existing is interoperability fail.
[00:21:14] People just won't adopt them. So you don't want to try and do more than you need to do. and so that's one reason why API APIs, even to being the other one is again, remember mech starts with a C it's a cloud computing standard, right? We play ultimately in somehow the same space as the de facto API standards that a global cloud platform provider does now, they don't do it standards, body.
[00:21:50] They just published their own standards, their own. Yeah, that's right. But that's the space we don't need to do more than that because that's all you need for [00:22:00] interoperability in the areas where we want it without interoperability was needed, which is in a case of mic two very distinct areas. One is between the application providers and the.
[00:22:13] Platform that supplications again, Iran in terms of how you explain it, certain services and the other one as an extension of what an, the fee does, Is in the lifecycle management of the applications. How do you do lifestyle? How does an operator who owns this edge cloud or operates it actually deploy these third party applications on the edge?
[00:22:38]So that's where we focused. that's where a standard was needed. Everything else is for others to define.
[00:22:45] Matt: Yeah. Now what if I'm an app developer and, for all kinds of reasons, I'm like, I don't know. I want to develop for one operator's cloud. and even if ETSI [00:23:00] MEC provides a mechanism and I don't know if it does, I'd be interested in knowing this present mechanism for federating cloud capacity across multiple
[00:23:11]Alex: It's a work item we're currently working
[00:23:14] Matt: Okay. So at some point I could make a decision that says, okay, I want to write a, I want to write an application that uses the MEK cloud standard. and that means that I can deploy my application on resources across multiple, theoretically across multiple operators and get some leverage, want me to write once deploy many.
[00:23:33]but there's also a model which says I actually want to provision. Azure instances are easy, two instances, and I want Microsoft and Amazon to figure out how to relate to the telcos, so that I don't have to use this new API I can use and hire developers that come from this gigantic pool, Tens of millions of people that the major cloud providers have assembled, as opposed to the, dozens of developers that maybe today know how to deploy MEC applications. How [00:24:00] So in that relationship between the cloud providers and the telcos there's technical interfaces, there's business interfaces, there's competitiveness.
[00:24:10]it's, is it a zero sum game between the cloud operators and the telcos, or
[00:24:15]for
[00:24:16] Alex: Close to it. So I've publicly said for a long time that in the final state of this MEC, Public edge computing game, there is presence from all most of the major cloud providers in most of the operator networks, because at the end of the day, you have small. Developers in a garage who are tied and committed to AWS and you want to capture them.
[00:24:51] There's 10,000 of them. You don't right. The whole game was, you don't know which ones to target. Cause you don't know which one's going to be the unicorn. So you just want to [00:25:00] capture them. And I think it's hard to. Capture them. And then in any other way than allowing AWS in the question. So in that sense, it's not a zero sum game.
[00:25:09]it's an own commodity. And you're going to do the same thing for Azure and an so awesome. And 10 cent in China, right? The question becomes how much of. Your business. Do you see to these guys? Do you see all of it? In which case, what a
[00:25:31] telcos are?
[00:25:32] Matt: how the telcos feel about that.
[00:25:34]Alex: then they are right. And they are going to be exactly where they were when the 4g game lost itself out, they were connectivity providers.
[00:25:43]or do you try and capture some of the value. That going up, the stack brings bringing your own cloud. And if you try and do that right, then you are back into a problem or, you can't have an AGM and a Vodafone and a Verizon [00:26:00] telephonic on a K DTI cloud. you can, and you will, but to a developer, they have to have it standardized set of KPIs.
[00:26:07] That's the role that ETSI MEC tries to fill. and I always tell people when I pitch. As the MEC, once you understand that, right? If telcos are going to succeed in having in capturing that value, they will have standards in a space where see my place. So it doesn't mean they have to adapt that's the max standards, but they will have to agree on a set of standards in that space.
[00:26:34]Our value is. We're the only standard right now that plays in that space. So it was so you can either go and form a new group and start from scratch, or you can go with ETSI MEC
[00:26:46]Matt: now, there have been some attempts, over the years of telcos attempting to offer something that looks like a cloud, for awhile, Verizon was buying data centers and things like that. And, outside of some of the really strange [00:27:00] examples like Rakuten, I can't think of a telco provider that's actually successfully built the developer program and attracted lots of developers.
[00:27:07] Can you.
[00:27:07]Alex: Yeah. telcos have let's put all the cultural thing aside, right in 4g, they had a fundamental problem. Developers wanted to talk to global companies that cross telco lines, because at the end of the day, if I'm a developer, I want to be able to capture all the users, which means that I need to be able to cross, in the U S right now it's three guys, four, if you count and dish.
[00:27:34]and where are the easy country? There's, some 50 in Europe and that's not counting tier three sensors. that's true. Telcos were always disadvantaged in that sense. the over the top players offered a way out, And they did it in a way where the telcos really didn't have the leverage to push back in the sense that the OTT [00:28:00] guys spoke directly to the consumer.
[00:28:03] Instead of, Hey, we're giving you this thing for free now. Go and yell at your teleco
[00:28:08]Matt: one could argue that those OTT applications are in fact what drove the next wave of
[00:28:13]Alex: of
[00:28:13] Matt: hasn't been, a total loss.
[00:28:15] Alex: right. I'm oversimplifying.
[00:28:20]Yes. But when you talk about this forge use story and how the telcos lost the yes. Verizon is doing all right. But when you, when they look at the revenues of an AWS, then they look at their own, they go, yeah. it's nice to have money, but the other guy's house has stands.
[00:28:41]And you could say, Hey, they weren't as aggressive. It's a different culture, right? I suspect you take all that aside and you just throw it out. The environment would have still forest the same final outcome. Cause they had no leverage when you go to five G specifically, because [00:29:00] applications that so many of the five G applications need the edge, an edge means location.
[00:29:06] They now have this leverage. Now what they're going to do with it, Suddenly would you're seeing is Amazon and Microsoft and lots of other people have to talk directly to telcos. They can no longer talk to the Teleco's customers. They have to talk to
[00:29:27] Matt: Yeah. in real value to, I look at like sample, the network slicing capability of five G. And if I was a developer and could wave my magic wand, I would love to be able to use all of my tooling on say Amazon to not only orchestrate my services out in edge locations that are in close proximity to the access network.
[00:29:52] Yes to provision a slice. and if you don't have a, if Amazon didn't have a business relationship with the service provider, there's no slice
[00:29:59] that's going [00:30:00] to
[00:30:00] Alex: Yes.
[00:30:01] Yes. So all of that, the foundation is different. Now that doesn't mean that the end result isn't going to wind up in the same place, dynamic because Yes. Yes. And the telcos cards are much better. and let's see what they're going to do with it.
[00:30:26]yeah, in that sense, it's not a
[00:30:28]Matt: I am too. I am enjoying watching it too. And I think you know it. Yeah. it's going to be, it's an interesting world. so one of the things I wanted to ask you about related to this is, what's some of the functionality that MEC exposes to a developer that I wouldn't have access to if MEC didn't exist, like what's unique.
[00:30:50] Alex: In two sentences or two phrases information about the underlying network or information that's only [00:31:00] available in the underlying network and the ability to steer traffic based on that information. Okay. So let me give you an example. very simple. For example, you are an application, let's say you're a game developer because that's what everybody talks about with Ash.
[00:31:16]And you've got this game and. In order to make it work. You really need an instance of some of your services that are on a right next to the users. Okay. So I take my phone and I start this game on my phone. And so now you get to go and I'm on Verizon. So you've got to spin up somewhere in Verizon's network.
[00:31:37] You guys spin up an instance. That's next to me. Now you could of course, ask my phone for the location and say, Hey, based on this location, There is a server that Verizon told me is located close by that. And so that's where I'm going to go and provision my instance. and that's people's thinking now, [00:32:00] There's pitfalls in that thinking one is what you really care about is my location on Verizon's network. And. If I'm connected to a macro tower versus some kind of a small cell can make a huge difference in a network location. Cause those network paths, depending on how the network is designed to telecom may not interconnect way in a coordinate work.
[00:32:27] So you spin up an edge distance, an edge, an instance, I may be a hundred meters from the macro tower, but if I'm not on it that the wrong place to spin it up. And you really don't have the, that information to give arises. other thing is
[00:32:41]Teleco's rightly Man, I'd be that willing to just expose what the topology of their edge centers looks like to every single application provider, which of course makes it difficult for an application provider to say, spin it up either. What you really want to do is you want to take my identity, which you can [00:33:00] easily get right from the networks and say, Hey.
[00:33:06] Yeah, like the unuser phone number or something and say, Hey, I've got a client with this ID, whether it's a phone number or something else. I need an edge instance spun up for this client per SLS that we have with them. Let the operator do that now in order to do that. You have to have, cause. the operator may know that there is a phone connected there. They have no idea what apps somewhere on it. So you have to have some agents, you have to have some my KPIs that you can quietly and find out where things are on and right. And then you have to say that, Hey yeah, for my traffic, when that instance gets spun up, this is how.
[00:33:48] This is the kind of traffic, right? So these are the five topples in the simplest case that I need terminated here. These are the five topples that you send back to your core network and out to the [00:34:00] internet. Again, not something that an operator would or should have wants to know about your application.
[00:34:05] They just want the instance to tell you to do that when it spun up on the platform. So those are the kinds of things that ETSI MEC enables
[00:34:12]Matt: I've seen a phrase, in some of the writings about ETSI MEC where they use the word contextualize or contextualization. And, I think I know what that means. but I'm super interested in how. Yeah, ETSI, MEC uses that and how they think about it, what it really
[00:34:30] Alex: all the actually I've never used
[00:34:32] that term because most people won't know what But. It's not about term, right? basically
[00:34:41] we allow you to understand the context of where your
[00:34:44] users are
[00:34:45] and where are the instances
[00:34:46] Matt: like you were saying, whether I'm attached to some macro tower with this latency or some small cell that's
[00:34:53]Alex: let's say you have an operator managed network in the mall, whether it's a Starbucks. It's super [00:35:00] important for you to do things right? To know whether your user's connected to Starbucks, to a wifi in a Starbucks, To your network, Or whether that user is connected to a macro base station, L not if it comes out of the same.
[00:35:16] Mad back then. so that context is super important and it's important for all kinds of reasons. not just deploying the instances, but Hey, what if you want to do targeted
[00:35:26] advertising
[00:35:27] while it's super important for you to know who's on
[00:35:30] Matt: And the operator can charge for that
[00:35:32] Alex: Yes,
[00:35:33] Matt: You want to target the Starbucks customers? Guess what?
[00:35:35] I have that information and I'd be happy to
[00:35:37] Alex: Yeah. And By the way, speaking of origin stories, right? One of the API APIs that we've standardized was actually originally defined way back. In a small cell forum, and then, Oh, I may standardized it. And we basically included it in that group of things that, that we do. and it's a, it was called zonal presence, API, and [00:36:00] all it really allows you to do is quiet and say, Hey, this device, what zone is it?
[00:36:06] Now you're going to ask me, what the hell does that do? it gives an operator the ability to create a zone that says all the Starbucks and central New Jersey. And now you can target your advertising to say, Hey, It's not that hit, that Starbucks hit that Starbucks. Hey, did I hit all the Starbucks in central New Jersey?
[00:36:27] Or did I miss too? I don't know. You can go to them and you can say hit all the Starbucks in central New Jersey. If you are the operator that owns that network and let them do that for the operator. Does that work well? and you're happy to pay something for having that contextual zone created for you.
[00:36:47] Because you're not paying somebody to sit in front
[00:36:52] of a spreadsheet and go, yeah, let me
[00:36:54] make sure that I'm
[00:36:55] still
[00:36:55] Matt: that's super interesting. So then, so it potentially becomes a relationship [00:37:00] where Starbucks can partially monetize their wifi is advertising space. Use that to offset the fees, the private LT or the private 5g or all the wifi that the telcos provider in an exchange, the telco provider can provide this, like add a placement contextualization.
[00:37:17] Alex: We were doing this cause I was in a small cell farm, So we were doing this back in the early 2010 and small cell forum was going to it was going to build on that and essentially stopped because at the MEC came along and we said, the right place to do it as a miss standards group, not, I'm not in an industry Alliance that really shouldn't.
[00:37:38] Promote that, Hey, we should use this standard, but that's not all right. But yeah, the, these are years. these ideas are not new.
[00:37:45] They're just becoming a reality
[00:37:47] Matt: Yeah, that's really interesting, shift gears a little bit. so one of the most interesting conversations around edge computing, although I find people, trying to shift the conversation, but certainly for most of the time that I've. [00:38:00] Then on edge computing latency has been one of the big discussions and clearly latency is important and latency is a function of, location, speed of light and number of network hops and more or less.
[00:38:09]but now there seems to be two different camps emerging. and I think partly it's because of where they come from and what their assets are, and that's what they need to argue. So if you're in the business, so a good example, and I don't mean to pick on Microsoft cause I'm a big fan of Microsoft, but Microsoft says we are 30 milliseconds from everybody on the planet. Whether
[00:38:28] that's true or not,
[00:38:29]Alex: wait 30 seconds. Most of the time or 30 seconds guaranteed. Are you willing to sign an SLA that says if I am at 31 milliseconds, more than 1% of the time, I'm going to pay you 2 million
[00:38:43] bucks every time
[00:38:44]Matt: that's okay. So that's a whole different subject, which is the SLS and the discreetness, because you're right. It's like jitter, which is the variance on the latency is way more important than the latency in a lot of cases. but what I'm trying to set up is, this very distinct [00:39:00] conversation that's emerging, that I'm just starting to notice now, which there's the camp of.
[00:39:04]30 milliseconds. And then there's a camp of like sub 10 sub five milliseconds and the folks that are in the 30 millisecond cap. And again, this may be because this is, your confirmation bias, right? You see the world, you see the use cases that you can actually support. they say, look, we're not seeing any demand for latencies, less than 30 milliseconds.
[00:39:23] So we're going to focus on. 30 milliseconds. And these, this is coming from the large colo providers and the large cloud providers and all this. And then there's a whole other group that say, no, it's sub five millisecond, So 10 milliseconds. When you look at some of the use cases around sub five minutes, they're like, Oh, it's going to be for like robotic lays.
[00:39:41] And it's going to be for autonomous driving. But you talk to the people that. Run million dollar robotic laves and autonomous driving. And they say, it's not going to happen. there's no way the safety control loop in a car or robotic Lei is going to be controlled by the network. And look, I'm not taking a position here.
[00:39:59]but I'm [00:40:00] interested if you're seeing those camps emerge. And if you personally, or if ETSI as an organization has started to form some hypotheses around this.
[00:40:10] Alex: At sea, no, because from an ETSI MEC point of view, it's irrelevant.
[00:40:14] Matt: I want to know your problem. I want to know your personal thing.
[00:40:17]Alex: We can tell you what a latency of an idea of an API is me personally, a lot of the high five G was driven by those low latencies. And that draw the presence of edge.
[00:40:28]Computing. And so many years ago, when five G conversation just started, I attended a talk by Gerhard HeartWise, who was a big early promoter out of the restaurant. And he was talking about tactile internet. So you have actual, tactile feedback on, let's say your screen, but the processing happens in the cloud.
[00:40:51]guess what? You'll need a couple of unions, single digit milliseconds, VR needs, September [00:41:00] industrial automation, things like vehicular automation and they still do, and they are going to come. The problem is those are not the low hanging fruit. And if you're going to invest, give bazillions of dollars in rolling out his edge cloud, you want some low hanging fruit that can allow you to recoup that investment right now, when you look at the low hanging fruit, I didn't hear it 30.
[00:41:30] I heard 20, but the reality is we're the 20th 30. It's not single digits. And. More importantly, it allows you to place it in your CEO. So what used to be your CEO is right. Your first line data centers, as opposed to having to go out and put it in a cell tower and where you don't, where you need it lower than that tends to be in large enterprises, which have their data centers.
[00:41:55]So it's not a talking head. You usually have an enterprise edge [00:42:00] already available. That's why I think the real industry, Hey, I need to go and make money. I need to go and sell. Certainly when I sell stuff for HBO or I attempt to solve, we talk about the 20 milliseconds stuff, right?
[00:42:18] Because that's where somebody can recoup their investment now, which means we can edge computing.
[00:42:28] Matt: what are some of the use cases that are in demand today that are in that 20 millisecond sweet spot?
[00:42:34]Alex: so it's really very use case dependent for each customer. but I can tell you what, we're a lot of the markets that we're addressing Mark is that have a lot of remote locations. applications that run at the remote locations and you can start doing,
[00:42:55]a national store that has outlets in every [00:43:00] single
[00:43:01] major moral right.
[00:43:02] Or seven yes. Something like that. So great thing for edge computing. And by the way, there, you can go sub 20, because if you can aggregate back at the telco, but you can also just talk about placing compute at the level at the location there that's operating managed. so latency is usually not an issue there.
[00:43:24]Initial kind of industrial automation applications. They tend to be monitoring stuff. So you do video processing and then you go back to the core again, latency right there. It's not even 20, there can just latency insensitive. the next wave that I am personally very interested in.
[00:43:48] Okay. Where I think latency will start coming into play, but 20 milliseconds should be a huge improvement of what we have now. Zoom, my kids are having virtual [00:44:00] classes on zoom. They are all within, there's 20 connections into a zoom session and it sucks because all these 20 connections go to a server somewhere they're all within two miles of each other.
[00:44:14] Matt: in the same network, you shouldn't have to send it back to some centralized point. You're absolutely
[00:44:18] Alex: and it used to be, you go and you talk to video conferencing people and you go, yeah. But when are you ever going to have a case where at least
[00:44:26] one person is not. And then
[00:44:30] Matt: yeah. That is
[00:44:31] Alex: are I'm
[00:44:31] Matt: really good, I got asked that question recently and I said, edge computing. Isn't going to make that big of improvement because somebody else's
[00:44:38] Alex: and you know what kept scratching my head. when would you ever have that?
[00:44:41] I bet. 80% of zoom usage today is hyper-local.
[00:44:46] Matt: Certainly some, yeah, I probably wouldn't go that high, but some percentage. Yeah, you're right. That is really interesting. and also, and I thought of another thing and that is, there's been a lot of research. I don't know if anybody's been commercialized where they do with video conferencing, some of the same things they do with video games, [00:45:00] where they predict the movement.
[00:45:03] Because you've got motion, right? Your head has inertia and things like that. And maybe they, and so they have some algorithms that are running locally. They're doing really sophisticated.
[00:45:11]so you're not having to send the full video image back and forth every fraction of a second.
[00:45:15]Alex: again, I don't know, but I suspect sub 20 is enough. And when you get to that sub 20, again, the differences it's not linear it's when can you move your point of presence for the application one network hop back. so if I have to stay right, let's talk about cellular.
[00:45:33] If I have to stay right at the cell tower. I don't really care whether the latency is five or seven or 10. If that forces me to stay at the cell tower, it forces me to stay at the cell tower. If I know that once it hits 18 milliseconds, I can move from the cell tower back to our CEO. Right or an aggregation point while that's a big difference, right?
[00:45:56] So it's not linear, it's a staff-wise function. and I think right now, the [00:46:00] applications that people are seeing are, Hey, you can, in most cases you can live
[00:46:07] Matt: Yeah.
[00:46:10] Alex: conditioned building
[00:46:11] Matt: Yeah. Yeah. Oh, so my company makes small air conditioned buildings with
[00:46:15] Alex: there's buildings.
[00:46:16] Matt: have to be big. Yes. But they are buildings. They are building, it's not a, it's not a street side cabinets.
[00:46:21] Alex: yet deploying under every cell tower. And I don't even know whether that's the plan
[00:46:25]Matt: no, and we don't actually think the cell tower is the right place to deploy because that's where the remote radio head is going to be. But you need to be on the other side of the baseband unit. So you're gonna be at an aggregation point. That is, 75 milliseconds from the radio head.
[00:46:39] Yeah. Yeah. But it's essentially the same thing. It's effectively at the cell talk,
[00:46:42] Alex: Yeah. that's what I mean at the cell
[00:46:42]Matt: what's
[00:46:42] Alex: Yeah. that's what I mean at the cell
[00:46:44]Matt: what's really interesting is, so if you take take all the politics out of this, you can take all of the, like fears of new technology and inertia and all this you think about. Okay. So I'm a company and I've got a factory and I've got a data center and I have it, people do.
[00:46:58] I really want to. [00:47:00] Own a data center. Do I really want to manage the data center? Do I really want to, can I do security as well as a cloud provider or a telco? there's all all these questions, and and my choice historically has been, cloud it's for security and latency.
[00:47:20] Alex: And you should make it clear to listen as security also includes data. to, or are not allowed to have that data, leave your premises.
[00:47:32] Matt: Yeah. Yeah. The data sovereignty issue. Yeah, absolutely. Absolutely. Yeah. and, but where I was going with this is I can see a world. So today, if I went to somebody, I said, look, I have a value proposition for you. get rid of the data center, I'll buy your data center.
[00:47:45]and I'll tear it to the ground. I'll take all your servers and I'll stick them in a data center. That's 30 kilometers away. sub one millisecond. You can come visit your stuff, there'll be people there. Or, there'll be remotely operator, there'll be an, conditioned area. [00:48:00]
[00:48:00]but you're going to have the reliability of a fiber line. I think a lot of it people would say, okay, that feels like it's onsite still. It's near prem, which is the word that I've heard people using. but then you say, okay, you trust your wifi network within your building? Shouldn't you also trust your 5g networking you're building, right?
[00:48:20]I can run your five G network from servers sitting right next to your equipment in this near prim location.
[00:48:27] Alex: Yes, I can give you that private
[00:48:29] Matt: Yeah. And in fact, there's even a crazier version, which to our earlier discussion may or may not happen is you don't even have to own those servers. It could be Microsoft or Amazon's potentially right.
[00:48:40] And so it's a really interesting
[00:48:42] Or GreenLake servers. That's right. Yeah. You buy them as a service
[00:48:46]or the operator service. That's right. That's right. So it's a, so it's a really interesting, we all talk about these on these near these low latency applications and mobile things, but there's a really interesting set of use cases, which is just replacing on-prem [00:49:00] with something that
[00:49:01] approximates on prem very closely, but gives you all the benefits of the cloud.
[00:49:07] So to
[00:49:07] Alex: absolutely. and that's some of the low hanging fruit that's out there. Unlike you said, 30 Millie's, especially if you can guarantee it and your, if you're an operator, you'll guarantee it If somebody wants to guarantee it's close enough to uncramp plus all the, let's not forget that operator data center runs critical infrastructure subject to a stack of regulations that thick.
[00:49:37] So they can probably give you security
[00:49:39]better than you could ever achieve with your little it department. If you're a small sized company.
[00:49:45] Matt: uptime guarantees as well. Cause they do that all day long.
[00:49:48] Alex: Yes. So there's definitely a real significant value prop there around them around enterprises.
[00:49:56]so yeah, it's, it's, like I said, it's a much [00:50:00] more interesting game than unfortunate game.
[00:50:01] Matt: Yeah. Okay. so I want to ask a couple last questions here. so if you look out into the future a little bit, and you think about where. all these ideas of where we could be, that we've talked about. there are of impediments or a set of things, dominoes that if we could nudge or topple, everything would move faster.
[00:50:20] What for you, what do you, if you could go and push it, push your finger against a few of the dominoes that would accelerate this whole new business model and edge computing and the deployment of hardware and the acceptance of standards and
[00:50:35] Alex: Show the operators the money. They want to see the money. And when I say that they don't want to see a strategy, talk about how much money the hyperscalers are making on the cloud and what portion of that market they can address and what that Tam for edge computing is wonderful imaginary world where everybody, every car is self-driving and [00:51:00] people are sitting in VR glasses right now.
[00:51:02] They want to show the money in six months, they want to see the money in six months. And, I think if you're showing them the money, it doesn't even have to be my experience. If you can show that the stuff pays for itself, they're going to be like, okay. It pays for itself. Amen. Because yes, there's the promise of all that stuff out there.
[00:51:24] So this is a way for me to start it that. Finding those business cases today is they are out there, but it's hard work because it is always on a case by case basis. You can't, there's no standard formula that I've ever seen that says, every single telco can do this at Southern boom. They make money and they put themselves in a right position where they, that they want to be in.
[00:51:51]you can certainly say. I'm just going to make my sites available for Cola Amazon, Google, and [00:52:00] AWS, and you're going to make money and that's not necessarily a bad decision. but in many cases, that's not what, that's not a direction that you want to go. I think that's the domino, the technology is emerging now.
[00:52:18] The other big, the technological domino that's out there, automation technologies. Those have to really, when you're talking about automating across a very, you're automating a cloud, which is highly distributed with very few small footprint at each
[00:52:42] Matt: Yeah. Yeah. it's a different class of problem. Yeah. Yeah. I agree. There's a lot of really interesting stuff that's emerging. but there's not yet, this dream orchestrator that you can handle the workload and a price you're willing to pay, and it's going to go find some edge server to run it on,
[00:52:58] Alex: Yes. And there was [00:53:00] a whole. It's almost there, but there's a whole host of little problems that, and I think in most cases, people are actually going about them. they're trying to solve it as one big problem, as opposed to partitioning it into lots of little problems and saying out of these little promise, 70% are actually solved. Let me just address. If I can address the other 30, I have a business case with something that's fairly
[00:53:23] Matt: Yeah, and I have to agree with that, but that's been my experience that in our business,
[00:53:29] Alex: That's a dig at some common friends of ours
[00:53:31] Matt: Yeah, exactly. in our business, we found that the projects that actually have a full end to end business attached to them are the ones that get traction ultimately. otherwise there's a lot of, sure.
[00:53:42] If you build it, we might come. but when there's a real end to end use case with, a disruptive business and a business model attached to it, that's when it really gets traction. So I could see that being the primary driver is, you get a dozen of those, that essentially pays for the infrastructure build out that everybody else gets to benefit from.
[00:53:59] That's [00:54:00] super interesting.
[00:54:01]Alex, this has been such a delightful conversation. you're certainly one of the guests that I would like to have back at some point and,
[00:54:07] see how the world's changed. Yeah. So I don't know when this episode is going to run, but so our audience knows this is we're in the middle of COVID and Alex and I are on a zoom call. and, yes, we'd both like to
[00:54:27] Alex: Google for us. there's two pages that come out. One is the portal. And the other one is they're both public, but there's an official webpage and a portal. And pretty sure my email in both of those, you can find me on LinkedIn. it's very easy. Alex arouse Nick. The Resnick is with a Z on with a K.
[00:54:56] So now see, and it's a Z anonymous, email [00:55:00] is Alex resnick@hpe.com and I have a book I'll buy a book, go buy my book.
[00:55:05] Matt: put a link to show what's the title of the book.
[00:55:07] Alex: the title is Multi-access edge computing inaction. and I have to say, when we wrote this book, we try to talk about the environment, right? the sort of the market, the technology environment, as opposed to a technical thing. So things that we
[00:55:25] Matt: Awesome. I'm a, I'm definitely gonna go read it. And I wish I had a chance to, I didn't even know you had written the book I found out earlier this week when I was prepping for the call.
[00:55:34] I love it. It's real. Can I get it on my Kindle?
[00:55:36] Alex: I don't know, ask
[00:55:37] Amazon
[00:55:44] and thank you, Matt, for having me. This was a lot of fun.
[00:55:47] Matt: great.