Podcast
Join Viktor, a proud nerd and seasoned entrepreneur, whose academic journey at Santa Clara University in Silicon Valley sparked a career marked by innovation and foresight. From his college days, Viktor embarked on an entrepreneurial path, beginning with YippieMove, a groundbreaking email migration service, and continuing with a series of bootstrapped ventures.
Links
Follow Me
Follow Me
Join Viktor, a proud nerd and seasoned entrepreneur, whose academic journey at Santa Clara University in Silicon Valley sparked a career marked by innovation and foresight. From his college days, Viktor embarked on an entrepreneurial path, beginning with YippieMove, a groundbreaking email migration service, and continuing with a series of bootstrapped ventures.
The Future of 5G and 4G: A Deep Dive with Guillaume Bélanger
Play On
10 MAR • 2024
42 mins
Share:
In this episode, I’m joined by Guillaume Bélanger from Canonical to explore the fascinating world of telecom innovation. Guillaume’s extensive experience in the sector offers unique insights into how open source is revolutionizing the industry.
We start by diving into groundbreaking projects like Magma and SD-Core that are reshaping private network management. What particularly caught my attention was Guillaume’s explanation of how these initiatives are transforming traditional network infrastructure. His breakdown of e-SIM technology and its potential to revolutionize device connectivity reveals just how much the landscape is changing.
The conversation gets especially interesting when we explore the technical aspects of 5G. Guillaume walks through the intricacies of core and radio networks, software-defined networking, and the practical challenges of setting up private 5G networks. His insights into the hardware requirements and regulatory considerations provide a comprehensive view of what it takes to deploy these systems.
I was particularly intrigued by our discussion about tech giants like Google and Amazon entering the telecom space. Guillaume’s perspective on how these companies are influencing the industry, combined with Canonical’s open source approach, highlights the evolving dynamics of modern telecommunications.
If you’re interested in network infrastructure, telecom innovation, or the future of connectivity, you’ll find plenty of practical insights here. Guillaume brings both deep technical knowledge and industry experience to the discussion, making complex telecom concepts accessible while maintaining their technical depth.
Transcript
Show/Hide Transcript
Welcome to this latest episode of Nerding up with Victor.
Today we're diving deep into the exciting world of 5G and 4G technologies with a special focus on open source innovation that is reshaping the telecom industry.
Our journey will take us through cutting edge projects like Magma and SD Core.
I'm thrilled to have a distinguished guest with me today, Guillaume Ballinger from Canonical.
Before joining Canonical, where he now leads to 4G and 5G management of private networks and projects, Guillaume hone his expertise at Bell.
His expertise and insights are invaluable in understanding how open source is influencing the evolution of 4G and 5G networks.
So gear up for an exciting conversation that promises to deepen your understanding of these technologies and the impact on the world around us.
So welcome to the show, Guillaume.
Very excited to have you on the show.
Thank you, Victor.
Thanks.
So did a quick intro.
Is there anything you would like to add to that intro to help listeners better understand who you are and perhaps the vantage point you're coming from?
Yes, of course.
So, yes, as you mentioned, I'm working at Canonical.
I've been here for two years now.
I'm engineering manager leading our 5G private mobile network offering.
So the way Canonical works typically is that we work closely with upstream community and we work at automation and packaging of that software to make it as easy as possible to deploy, but more importantly to conduct all the lifecycle operation of that software.
So here I'm talking about scaling, upgrading and integrating with observability and other necessary components that you may have in your infrastructure.
And I'm happy to answer any other question you may have.
Perfect.
Let's dive into the whole packaging side of things later on because I think that's kind of interesting and I kind of want to dive into that as well.
But the reason why I wanted to do an episode on 5G and 4G and LT and all these technologies because I, I think we're at almost like a pivotal moment right now with the evolution of 5G, where I think 5G might possibly be the new WI Fi if you fast forward 10 years out.
And I think the big game changer, at least for me.
And what I got really excited about is with esims, because with esims you're no longer locked into a singular carrier.
So.
So you can have three.
I had, I've had three or more three sims in my phone for a while and you can now use mvno.
So mobile virtual network operator that can serve a specific niche.
They don't have to have a countrywide coverage.
They can have super good coverage in a small area you're in perhaps, or even a coffee shop or so on.
So I'm just one of the pastor and say like, does this kind of.
How do you see this?
Is this something that you would agree with as a thesis or how would you basically see this world shaping up?
Yeah, well, it's hard to tell.
It's always hard to predict the future in those areas.
5G and Wi Fi at the moment serve different purposes and are very much a set of different technologies.
And 5G, it's a set of spectrum.
So here if we talk about the air aspect of it and each country has their own ways to acquire part of that spectrum, and it costs a lot of money for any given company to have access to that spectrum.
There's also a set of licenses that are required if you are to implement your own 5G radios that you would have to pay to company like Qualcomm.
At the end of the day, this means that operating 5G is much more expensive than operating WI fi.
So this is something that has to change over time if you want to get closer and closer to WI Fi in terms of cost.
That being said, I mean, they both serve.
There's also lots of commonalities between the two.
At the end of the day, the goal is to go have access to the Internet and most people don't really care about how they gain that access as long as it's the cheaper and reliable option.
So it's definitely a possibility.
But I see a set of things that need to happen in order for us to be there.
Yeah, I mean the legal part is obviously a massive part of 5G in general.
I think in the US you is probably leading the way with the CBRS spectrum where you can deploy without having these licenses.
I'm not sure you want to speak a bit about that and just enlighten people on what that means and how that kind of what that means for someone who wants to deploy a 5G router.
5G private 5G, for instance.
Yeah.
So in the US, CBRS stands for citizens Broadband Radio Service.
So it's a somewhat new service that the US is, has been starting.
It's around five years old, but very much implementation is starting in the past two years.
And the main idea, so I spoke about spectrum earlier and the main idea here is that if you have your own private mobile network company, you don't have to purchase spectrum.
That's a huge deal because if you look at spectrum costs in the US it's in the hundreds of millions of dollars for just a small band for a given geography.
So most people, most company cannot even dream of purchasing a part of that spectrum.
And the idea of CBRS is that you can use part of that spectrum without purchasing it.
So the way it works is you have companies that work with the FCC in the us that's the regulatory body that owns those processes.
And those companies are.
So Google is one of them, for example.
And they have a service called a spectrum access service.
And if you want to use those bands, you have to integrate with this service and they will tell you exactly which band you are allowed to use and when and then you can integrate with that.
This is groundbreaking in the sense that it is completely new.
It's the first and only country as far as I know that has such a system.
And the main outcome hopefully will be that a reduced cost in operating 5G.
This is huge.
I don't know if other countries are going to follow through.
In Europe there's been some bans that have been allocated to specific companies.
I think Daimler and a bunch of others did purchase some small part of the spectrum, but there's no such system where anybody, any company can just operate their own 5G network.
That's really exciting.
Yeah, I think there have been some pilots.
I think Sweden among other countries had some kind of pilots where you could apply for very small building level spectrums.
So the right to use the spectrum within a given building or a campus.
But yeah, it's not something you could deploy across a country for instance, but they will be like a much reduced price.
That obviously and kind of fits into the narrative that maybe your office would offer a, or college would offer a 5G service rather than use WI fi because those would be far cheaper and far more efficient to achieve.
So let's dive in a little more to like the status quo before we dive into the open source world.
I'm curious about.
You talk about these extreme costs for the spectrum, but the spectrum is one part of it.
Important part obviously.
But talk a bit more to me about the hardware setup.
How does a AT&T or a bell or insert blank carrier.
How do their infrastructure look like for the end user or like end to end really?
Yeah, sure.
So when we talk about 5G we talk about two different parts.
We talk about what is called the core network and then the radio network.
I'll start with the core network.
So the core network is the part that most people typically don't know about.
It's Not a well known aspect of 5G.
And this core network is very much the equipment and logic that makes your 5G network function.
So in this part of the network you have all the logic related to authentication, authorization, charging policies.
So it is more or less software that runs on a computer.
So in terms of hardware, you need computers, it's very much as simple as that.
As we get closer to the radios, one component of that core network is called the userplane function.
So this one is the component part of the core network through which user traffic goes through.
So this one, when you make any usage of your cell phone, the packets that go out of your phone will go through this function.
So this one, typically you would have a bit more specialized hardware in order to minimize the latency and time that packets need to be processed in this component.
But for the rest of the core network, it's just generic computers.
So in the user space, that's where the traditional telco vendors are, like Ericsson, Nokia, Huawei, that's where these guys have a significant market share.
Right.
And then I guess the control plane, I guess is the other side.
Is that fully standardized or like every kind of carrier roll their own system or how does that look right now?
Okay, so historically it used to be that companies, the examples that you mentioned would sell you physical hardware to install.
So they would call a company like BT or Bell Canada and they would install those boxes, physical boxes there.
Now with 5G, the code itself, the software itself is based on specifications written by 3Gpp.
So what Nokia, Ericsson and others do is they will implement those standards.
So it is fairly standard in that sense that they have to.
Well, they don't have to, but they follow those standards.
What changes is even though they do write implementation of such standards, they also have other integrations that are not part of these standards.
An example of this is what they call their network management system.
So 5G is a set of different functions, something like 12 different network functions.
If you want to configure those, you would typically need either a user interface or a set of APIs that to do so.
So those companies will sell you this network management function and that's how they make money.
At the end of the day, with those integrations, you don't just buy a generic network function, you buy a platform that comes with monitoring, that comes with a nice UI to configure it.
And more importantly, it comes with 24 hours 7 support from them.
If your network crashes, you have somebody to call for help.
Right?
Okay.
Okay.
So that's the status quo.
Before we dive into kind of the nitty gritty details of what this all means, I'm curious about going through some different type terminologies that some of them already been mentioned.
But just for clarity before we dive into the details.
So we already covered 5G and the fact that it's actually a family of protocols rather than a singular protocol.
Right.
So we can dive into that a little bit.
Do you want to a few words about what those standards are?
Because you already alluded to the fact that there's an IoT version, there is a regular mobile version, but then there are other families as well.
Right.
For other use cases, do you want to zoom in a little bit on that?
On 5G specifically?
Yes.
Yes.
Let's start with the 5G part.
Yeah, sure.
So I mean, 5G, I mentioned earlier the core network.
So like 5G is a set of network functions and communication protocol that those network functions should follow to communicate together.
So that's for the core network.
5G is also the set of frequency on the radio side and the protocols on the radio side that have to be followed between the radio and the modem on your cell phone.
So 5G is all of that.
So it's more than just one specification.
It's a set of specifications that companies and implementations have to follow.
Yeah.
So you have.
Yeah, okay, that makes sense.
Let's talk more about the specific ones because I think one thing I didn't know about was particular about the lower consumer cases, like IoT versions for 5G, for instance, how widely.
I know there's a standard for it.
Is that something that's widely deployed today or is that something that is just a theory or a protocol designed by gsma?
I presume it is.
Or how is that actually.
Yeah, do you want to talk a bit about that?
Because I'm curious about those use cases.
The truth is I'm not familiar with this.
I'm not sure what's the IoT specific standard for 5G.
So you may be.
Yeah, there's a low energy concept for 5G essentially to compete more with Lora for lower energy consumption.
So low bandwidth, higher range and low energy consumption.
But yeah, that's.
We can skip over that part.
That's fine.
4G and LTE are obviously different generations and different type of end user communication methods to the towers.
Do you want to differentiate between the two of them and talk a bit more about what they are and how they are different?
Yeah.
So for each of the generations of wireless communication, there is A set of standards that define both the core network and the radio network.
4G had its own set of components that are part of a mobile network.
And LTE was an evolution to 4G that was compatible with 4G that introduced some new components.
For example, one of those components was called the epc, the Evolved Packet Core.
So it was an evolution of 4G.
5G itself is a completely new design, a completely new set of functions and functionality part of that core network.
And if you were to implement 5G, the core part, you would have to do it from scratch.
Okay, so lte is basically 4G 4.5G essentially.
It's in between 4G and 5G essentially for higher throughput.
Got it.
Oh yeah.
Cool.
One thing that pops up a lot when doing research about this, and it's something that spans all network technologies.
Sdn, like software defined, networking.
That's a big concept.
How far is that being pushed on this?
Because I know on switch level, like on server side, that's kind of commonplace these days.
But how far is that deployed and used in this context here?
Yeah, so as I mentioned earlier, it used to be the case that you were to purchase a physical box from a company that would take the role of a given network function.
If we talk about 5G now, what you are purchasing from those companies is essentially software that will install a generic hardware.
So you would very much buy a server from a company like hp, Dell or whatever server company you prefer and install your 5G network functions from Cisco, Nokia or whoever on those physical boxes.
In this sense, it is software.
Now, not all of those component of the components that are part of 5G network are only software.
Some of them need to be, like I said, hardware accelerated earlier.
So software defined is very much like a trend that instead of buying physical boxes for something, you write software and you install it on boxes for routers and switches.
It is still the case that physical switch is faster than a software switch.
But depending on your use case, that may be fine.
It really depends on what you're trying to achieve.
How about SDR software defined radio?
Is that something that enter the space as well with like where you can redefine the spectrums basically in software using sophisticated silicon.
Really?
Right.
Or sophisticated hardware.
Yeah, yeah.
That's getting more and more popularity.
I mean, even if you take like Qualcomm's offering, in the world of radio, it is very much software defined.
So for the world of 5G, there is something called open ran, where it is a set of specifications so that you would implement so you would write code that satisfies those implementations and you would.
And that code can run on generic hardware.
So you can have an intel chip or you could have a chip that is actually made specifically for that.
For that purpose.
And yes, 34 radio is a trend in the industry where you can write code on that radio.
And there's more and more like smaller projects that are.
You could yourself buy such a radio and write code on it for yourself.
There's a project, for example, called Bladerf is exactly in this space where you buy a radio for something like 500 bucks and you use their project for GitHub and install it.
It's like a hack orf similar thing for radio.
Got it.
Okay, cool.
One thing that you've mentioned a few times now is that it's becoming standardized and it's like it's.
You're running on commodity PCs now.
And obviously that is a nice transition over to the software side of things.
And there are two projects that I kind of want to do a kind of talk a bit about.
One of them is Magma and that one is SD Core.
Magma, if I'm mistaken, probably came out of Meta and has had a quick rise and a quick fall from what I gathered.
Do you want to speak a bit about Magma?
How it's what it accomplished and why it failed in the end.
Really?
Yeah, sure.
So yeah, Magma is.
I mean it's still out there.
It's a 4G and 5G private mobile network that was led by Facebook at the time and at some point Meta.
So it started at Facebook or actually was announced by Facebook connectivity in 2013.
But the first commits showed up on GitHub in 2019 and most of those commits were made by Facebook employees.
So this was very much led by Facebook.
Now, in terms of why they were spending any effort on it's generally to increase connectivity in the world.
And that next part is speculation so that more people can install Facebook on their phone.
So now in 2022, there's been a huge round of layoffs at Meta and they completely reshuffled the projects and focuses of the company and the whole Facebook connectivity.
So not just Magma, but every project under that umbrella stopped being funded.
Now, to facilitate the transition, they did pay a firm in Germany to help with this transition.
So they would continue to work for a specific amount of time for I think it was one year.
But that funding stopped at some point in March 2023.
And at this point it is a community led project in the sense that Facebook doesn't spend any engineering effort on it.
But the truth is the amount of commit has drastically decreased.
Right.
I mean, now, what brought this project to this situation?
It's always hard to know for sure.
I'm not behind the scenes at Facebook, but one thing that has to be said is that Facebook didn't make any money directly from this project.
So when things go rough and you need to find projects to cut, makes sense from a business point of view to cut these categories of projects.
Yeah, that makes lots of.
Yeah, guys.
Sorry.
Yeah.
And in terms of growth, I mean, Magma didn't.
There's been a lot of excitement around Magma when it started, but for the last couple of years, I mean, the amount of users and companies implementing it didn't grow that much.
So I suspect they felt that's a good time to let go.
If I'm not mistaken, Amazon's 5G offering did use Magma at one point, I believe.
Their private 5G offering.
Maybe you have more insight to me than me on that one.
Yeah, well, supposedly all of that is not public, so I don't know for sure what they did use or they're currently using.
This could have been part of a exploratory phase or actually implementation.
Yeah, I don't know more.
Okay, fair enough.
Yeah.
Because obviously had they pushed ahead there, they could obviously have the manpower to and revenue stream to actually maintain such project.
Right.
They would have much bigger incentives than Facebook.
Cool.
Let's talk about SD Core, which is the new kid on the block, it sounds like.
Well, maybe not new, but what has taken over a lot of the traction from Magma from what I gathered.
Yeah, well, so Magma was a.
Or it still is a 4G and 5G private mobile network.
So it is a set of technologies that are bundled in a package as of SD Core.
And also another implementation that comes from an European organization called OAI Open Air Interface.
Those two projects aim to be much closer to the actual 5G standards.
So those are actually like, you read the 5G specs and you implement them in code.
They don't do much more than that.
This is both a pro and a con.
It's a pro in the sense that it is really following the specs.
You can be pretty confident that it is compatible with anything else that is also following the specs.
But Magma was trying to be more than that in the sense that it also contained a nice user interface.
An integration with.
Earlier we spoke about cbrs, Magma came with integration with Spectrum Access Systems, all of those that are not necessarily part of a 5G specification.
Now, SD Core is a project led by the Open Networking foundation and it is an implementation of 5G core specification.
So all of the network functions that are part of 5G Core.
So those 12 or so network functions are being implemented in GO code.
And us like Canonical as a company, we package and distribute those as well.
Okay, so what in terms of deploying your own?
Because I know they were like a juju deployment mechanism for.
For Magma for instance, which I presume is kind of dead in the water now, or it might go into maintenance mode, I would imagine.
Has the offering around SD Core replaced that from the Canonicals roadmap?
In terms of what you guys are pushing on your end or how do you guys see the world there?
Yeah, yeah, we're in the process of doing that.
So we so did.
This is the type of project that takes a while, but we are in the process of doing this.
We are also involved in the actual upstream project.
So we work with the upstream community to improve the actual software that is distributed and we also package it and distribute it to make it as easy as possible to operate.
So I mean, if you are to follow the upstream deployment of the project, this is done through helm charts on Kubernetes and it works.
What we are trying to offer here is a more universal way to deploy SD Core.
For example, in the option project, there's only one distribution of K8s that is part of this deployment script.
In our case, we have compatibility with any cloud kubernetes.
So if you're using AWS's Kubernetes offering or if you're using your own personal microkits on your laptop, both work with this approach.
If you want to scale and upgrade, we offer those paths as well.
Okay, so speak to me a bit more about like if I wanted to deploy my own.
Let's assume for a second that I am in the us because it makes decisions a lot easier around cbrs and you have your own.
You can use a CBRS spectrum.
What does it take for me to set up my own 5G network at home from scratch if I wanted to run my own?
Imagine for a second I live in the middle of nowhere, I have Starlink, but I have zero connectivity.
So I want to set up my own 5G network at home.
Walk me through how that would look in terms of end to end.
Yeah, sure.
So from the core network part of the story, you would need a computer.
And if it is just for yourself, you would very much be okay with something like Intel NUC or Just a really small computer that has a relatively modern CPU in it.
And this computer would run all of your 5G core network.
Now you would also.
So on this computer you would install a, ideally a Linux distribution and you would install kubernetes on top of it.
And you would use something like a juju but not forced.
You could use also HELM or something else to both deploy and operate your 5G network.
So all you need is this one computer on which you install this software.
That's for the core network.
Now you also need a radio that you would buy from a radio vendor.
Those tend to be a bit more expensive they that were talking about cost for 5G a bit earlier you would have to spend between $1,000 and $20,000 for such a 5G radios.
Those are quite expensive depending on if it is outdoor, indoor and the specific technologies that are in this radio.
That could vary.
But those are essentially the two things that you need.
The core network is going to be pretty cheap in this story very much.
You're going to spend most of your budget this radio equipment.
Right.
And then okay, so I have the software stack installed.
I always need to connect my phone, right.
So walk me through the process of I guess issuing ESIMS or SIMS bootstrapping.
Like how does that process look like if you want to run your own infrastructure.
Yeah, for physical sim you would need to have a SIEM writer.
So a small piece of equipment that can write SIMS SIM cards.
And on your core network you would have to register.
Okay, I have this new SIM for this new user.
This is the set of policy that I want this user to follow.
For example, I could give you a 20Mbps speed with a certain priority.
So there's a concept of priority in 5G and 4G which decides like okay, well when this person and this person talk, this one has priority where that type of stuff.
And so yeah, you would need to write that SIM and register it in your core network and install this SIM in your phone and you should be good to go.
That's for sim.
For esim, the story is a bit more complicated.
I haven't worked much with ESIMS yet.
This is still to go, still to come.
But it depends from what I understand on what is your cell phone.
For example, the process of becoming a carrier of ESIM on an iPhone versus on Android phone is different with Apple.
As far as I understand, you need to be somewhat of a partner with them.
So I cannot go on my own and become a 5G carrier for some Android it seems a bit more straightforward, but like I said, I haven't gone through this process for either of those.
So I'm not certain this process.
Okay.
And then I guess you have some kind of serial number that.
From the SIM card, I presume that.
Or how's like what's the cryptography of this Vahan?
Like when you write that SIM card, like is it a.
Is it a private key that goes on there and the public key that you register?
Or like what does the cryptography side look.
Look like?
Correct, it's a public key, publicly private key system where you would.
Yeah, it's exactly that at the end of the day.
And whenever you register that SIM card on your core network, you would register its public key.
So both sides have a public and private key.
Right.
So it's very much.
It's not so different from any HTTPs communication.
Right.
So it's basically like a small local PKI essentially that you can.
Okay, so you.
And are there like clis for this or like is it all API based or the uis for like if you wanted to build this yourself, like what does that look like in terms of experience?
So on the core side to add and register sim.
So this part is very much not part of the 5G specification.
So each company implements their own approach to subscriber management.
But if we talk about our distribution of SD core, like we write our own network management system where you can create network slices, create network device groups and create subscribers.
And so it is a front end written in NEXT JS that people can use to configure their network.
So yeah, it is a pretty nice front end with this purpose.
Now the process of writing SIM cards itself.
There's companies that sell their own SIM riders.
I don't know of any nice open source project there.
There is some really smaller projects.
There's something called PYSIM out there which comes with a CLI to help you write sim.
If you have a SIM writer.
I haven't worked with it yet.
This part is.
I think people will tend to go directly through esim, but there needs to be a bit more of an open source approach to it, which likely will go through Android before Apple.
Right.
So it sounds like it's not super straightforward given the state of the project today.
To build your own 5G network in your house.
It requires obviously hardware, but that's inevitable I guess for the time being.
Well, forever I guess.
But cost wise, I meant.
But there are like how do you.
Like, do you.
Because it is like is it kind of proof of concept level, quality or is it like stable production level?
How do you rate the stability or maturity I guess of this?
Yeah, today I would call it like an mvp.
You could deploy it, you can use it to run subscribers.
I would discourage you from running high super critical load on it.
Like I would very much discourage you from using 911 calls.
And this network, it's still in early stage.
We started working with SD core a couple of months ago.
So it's still very early development phase.
But as we go through user validation we will gain feedback and improve the product for sure.
Cool.
So we talked about before we turn the table to some other things later on, but before we do, I'm curious about.
We talk about data routing and that's obviously very similar to network routing in general.
How about phone call routing?
How does that work?
Because that's usually tied to your SIM card and you usually get identity there.
How does that whole interface with the rest of the world?
I have my own interface.
I can run network data.
But what if one actually make phone calls?
How do I actually register that and how would that actually work?
Yeah, you would need to.
So here we only touch about the SDCore really does only the data part.
So we haven't yet worked on this phone aspect of it.
Again this is the type of projects that is going to come up fairly soon.
But right now you would need to integrate with a mobile network operators system for managing those phone calls.
So yeah, coming up, but not currently possible.
Got it.
Okay.
So obviously this is very much in the research phase of where you are at canonical, how you see and obviously strategic bet obviously for the future.
Talk to me a bit more about how you see private 5G being deployed in the conversation you guys are having with and who's actually interested in this is this companies and campuses that are mostly interested in these conversations or other conversations with mobile carriers that are looking to deploy this as well.
Yeah, those are two sides of the coin for us in terms of private 5G.
So for companies there's a bunch of factories that are Interested in having 5G mines, earth rigs anywhere that is typically poorly served or that you want isolation from the rest of the world in your 5G network.
Those are all potential customers and they're good use cases for us in the sense that in terms of it's not running critical infrastructure like the.
I mean it's not a hospital.
But that being said, hospital are also good use cases for private 5G.
And then there's private, then there's mobile networks for Mobile network operators.
So a company like BT or Bell, they receive requests from companies that are interested in private mobile networks.
Most enterprises don't necessarily have the skill in house to operate such private mobile networks, which of course makes sense.
But they want 5G connectivity, so they would call a company like Bell and ask them to run that for them.
So it is a case where Bell Canada would spend much less money by running a private mobile network than running their more country wide grade 5G network.
So those two cases are possible for private mobile network.
Right?
And yeah, that all makes sense.
And in terms of architecture, the actual antenna piece I guess is a network connected device.
Can that then just talk to some cloud interface?
Or is latency so important that you need to have all the other pieces on site as well?
Or how does that infrastructure look like for something like this?
So for some part of the system, latency is super important.
So the antenna will typically talk to the component I talked earlier about called the user plane function.
And this user plane function will be close to those antennas.
Physically close, right?
Not necessarily.
And then that physically close number depends on your use case.
You may want it to be in the same building, you may want it to be in the same city.
That depends on your latency requirements.
And then the rest of the core network is not going to impact latency as much.
So the rest of the core network is going to be responsible to tell the user plane function and the radio whether or not the user is allowed to use the network.
And they're going to manage the session, but they're not going to impact latency as much.
So part of it can just be on a cloud system.
So most of the core network can run on AWS or Google Cloud.
The user plane function will typically run closer, like maybe in your factory.
And then you can have a couple of antennas in your factory that are linked through that userplane function with either fiber or ethernet cable, depending on the load that requires.
Right.
Do you see a world where aws, Google, all these guys are essentially going to go head to head and compete with the traditional telcos because they have the infrastructure always run the control plane better so than the telcos would and have far higher scale and would do it more effectively and cheaply.
Is that a scenario you see playing out or how do you see the world there?
Yeah, well, it's a possibility.
I know Google tried it a little bit.
For some part of it is possible.
I mean most of the telcos are now going through a phase where they are moving a lot of Their private clouds to the public cloud.
So clearly the public cloud has something that they don't currently have.
However, a big part of operating a telco is laying fiber, cutting trees and installing radios.
And I don't see Google going in that direction.
So most of that part of the work, I still see a place for Taco, but for running servers.
Definitely a place for the cloud.
Yeah.
And that's kind of like you said, Google played with that.
I think Amazon had something they called curbside something, a service that I think they were using for some of their.
Was it Alexa?
They use some backend as well.
I think they use some of their devices to create like a semi telco, but I think it was based on lora rather than 5G, I think.
I'm not sure you've seen anything about that project.
I haven't seen it, but definitely, I mean they're getting more and more in the space of telcos, they clearly all want the telco business.
This is billions of dollars worth of money, but they are getting it.
I mean, if you look at a company like Google, they are currently working on a project called Nephio, which is a open source automation project and it's fully open source, but they work on these kind of projects to attract the telco so that they run their workload on Google cloud instead of another cloud or on their own private clouds.
Tacos still have currently huge clouds to maintain, so thousands and thousands of servers that they run throughout the country to run those 5G network services.
And clouds are more than happy to take that workload.
Yeah, absolutely.
I can definitely imagine that.
And I think that's particularly with the commoditization of the infrastructure and if you can use things like.
If I'm mistaken, I believe Google's home hubs, I believe they have a level of SDR in them where they can probably not do 5G perhaps, but you can envision generation three or four out where they would have support for this built into their home hubs.
Right.
And then it's definitely an interesting development.
I think this is super interesting and I think we've covered a lot of grounds and I'm not sure there's anything else you think it's worth bringing up in terms of the development and how you guys are working on and what people can learn more and how people can help develop and test this out themselves.
Yeah, so most of the projects we talked about are open source.
So both enterprises, but also individual can work on those.
They tend to be really fun for university projects.
The sense that people can work on their own during, for example, their master's degree, add some feature and then having this feature merge at the end.
Those are exciting for this purpose.
And on the other side of things, what as a company we're trying to do is bring those to a step that they become actual products where you are confident to run high critical workloads.
So, I mean, to learn more, you can go on the ONF website, you can look at Linux foundation networking, you can look at Open Air interface, or if you really want it, you can read the actual 5G specs.
But I suspect that's not free.
Yeah.
And how far do you think Canonical, or you guys in general are from having this as a commodity item in the sense that just like you would deploy Kubernetes, you could Deploy your own 5G stack.
How far do you think that is away?
Very close.
I mean, to run data, you can already do it.
So this is gonna.
It is a work in progress and it's never gonna be fully finished, but yeah, it's a matter of months, not a matter of years.
That's amazing.
Yeah, I'm definitely looking forward to give that a go.
And hopefully the legislative side of things in Europe will make it easier.
I'm not sure how things are in Canada, but I think it's in the UK at least.
It's still difficult to get those approvals.
It's similar and it's very much similar across the world.
Perfect.
This has been super insightful for me, Guillaume.
I'm very happy for you to have joined me on the show and yeah, thank you so much for coming there and have a good one.
Thank you so much.
Thank you very much.
Cheers.
Bye.
Found an error or typo? File PR against this file or the transcript.