Telco Infrastructure Evolution - Automation, AI and Containerisation Explained

Matt Hurst, Senior Director of Strategic Business Architecture at Ribbon, joins Tom Arbuthnot to discuss the significant changes occurring in telco infrastructure with AI, analytics, and automation. Matt clearly explains how Automation, Containerisation and AI work and apply in telco and also large enterprise infrastructure.

The telecommunications industry is embracing containerisation and automation to address competitive pressures, security requirements, and operational efficiency challenges. Traditional hardware deployment models are being replaced by flexible, cloud-native approaches that enable rapid scaling and automated management.

• Containerisation breaks network functions into modular components that scale dynamically based on demand
• Automation reduces manual patching, testing, and configuration whilst improving security compliance
• AI-powered chatbots guide junior engineers through complex troubleshooting using integrated documentation
• Infrastructure as code creates single source of truth for network configurations across multiple sites
• Adoption is being driven by security operations requirements and specialised skills shortages

Thanks to Ribbon, this episode’s sponsor, for their continued support. 

Transcript

Matt Hurst: The operations teams are having to do more with less people and more with perhaps people that aren't super skilled in a particular product or a particular technology. And that's where another piece of like the AI comes in is, you know, we, we give the ability, you can build a, like a chat bot and then it can interface with trouble ticketing systems, but it can give like an interface to a junior engineer that they can interact with.

Tom Arbuthot: Hi, and welcome back to the Teams Insider Podcast. This week we are talking big telco infrastructure. That might be in operators, but it might also be in large enterprises. I'm talking to Matt Hurst at Ribbon around automation, containerization, AI, and how all of that is impacting the choices of telecom providers and their infrastructure.

Really interesting conversation. Hopefully really useful to get a handle on containerization and automation and AI in Telco and how it's impacting that space. Hope you find the conversation interesting. Many thanks to Matt for joining the podcast and also many thanks to Ribbon for all their support of Empowering.Cloud.

On with the show. Hi Everybody. Welcome back to the Teams Insider Podcast. This week we're getting into a bit more of an infrastructure conversation. I'm really looking forward to having it. We're gonna talk automation, containerization in the kind of telco infrastructure space, but also in the large enterprise as well.

Um, got Matt Hurst. Matt is a familiar face to the community, but for those that don't know, you, Matt, maybe you could give an intro and then your new newer role as well. 

Matt Hurst: Yeah, absolutely. Uh, thanks Tom. Great to be here. So Matt Hurst, I'm based outta the UK and I'm senior director for our, uh, AI and automation practice now.

Uh, so previously been working as you know, a lot on the, the Teams and the UC, UCaaS infrastructure. But yeah, my newer role is really around automation, um, and the implications of of AI and how we've helped the telco and help our enterprise customers on that journey really. 

Tom Arbuthot: Awesome. It is a really exciting space.

Obviously a, a lot's been going on in AI automation in general. Right. Like a massive steps forward. Yeah. Um, and it's really interesting to see how that is coming to Telco and, and how much it's impacting and how fast and what people are doing. I mean, telco traditionally not being the fastest moving industry for honest.

Yeah. Like, it's, it's, it's, it's mission critical. It's, you know, really carefully considered it's multi-year projects. But things are changing and particularly influenced, I gather by the kind of software containerization, Kubernetes type, you know, cloud world. 

Matt Hurst: Yeah, absolutely. You're right. I mean, the telco world has traditionally been very, um, static in the way it develops and evolves.

Um, but I think there is, there has been a growing trend amongst telcos and large enterprises, as you say, as well to embrace, uh, containerization in particular. Um, for a number of things that the benefits we'll talk about, but also then automation. How can we remove some of the more mundane tasks that we undertake as a, an, an organization?

How can we make things more efficient? Um, and how can we also keep up to date in a, a much more rapid way? You know, particularly telcos. When you, when you look at network component upgrades and things, it can be a very lengthy cycle. You know, you're talking between six to 18 months to get an entire network, uh, software upgrade done, for example.

How can we improve that? How can we improve our customer service and satisfaction levels? That's the key driver really. 

Tom Arbuthot: Yeah. And two things I'm seeing, like I talk to a lot of the telcos, particularly the operator connect telcos and it's, it's highly competitive. Yeah, right. There's lots of people in that space.

There's lots of moving parts. People want to get into new countries, so some of that need for agility to compete. And it's not just OC, it's the other UCaaS as well. Like, like the need to be compatible with those. They need to compete. And then security is the other one I'm seeing a lot. Obviously security is a hot topic for everybody, right?

But like. These infrastructures are being attacked daily, hourly, all the time. So like, like, like needing to be patched and current. Not being able to run on the stable firmware from 2024 is just, it's not so plausible anymore. 

Matt Hurst: Yeah, no, absolutely. Security is top of mind for all these organizations and, and the first point you hit on there is, is critical for telco operators.

How can they remain relevant with their customers when their, their margins, their traditional voice, minute margins are getting eroded by some of the OTT players. Um, and a lot of people sort of say, you know, voice is dead. It's like, voice is not dead. It's what we do with voice and how we can evolve that further for the telcos and how can they build new services.

Um, and including AI. You know, AI is a tagline, but I think there are some real business use cases starting to emerge from this. And it's a case of how can telcos adopt that? Quickly versus the more traditional, very static manual process that would, that would just never get off the ground. And that's where the automation comes in as a key component.

Tom Arbuthot: Awesome. So maybe we could kind of take it back to the fundamentals. Yeah. I mean, uh, we say containerization, uh, containerization. We say automation, we say AI. Can you kind of set the foundations of what even containerization is? 

Matt Hurst: Yeah, absolutely. It's, yeah. Let's break it down. 'cause I think there's a lot of, um, terminology if you throw about and thinking everyone understands what they are, but let's, let's take it back a step.

Tom Arbuthot: Oh, I'm, I'm, I'm on the edge of my understanding that, so you can educate me today. 

Matt Hurst: Yeah. So first of all, I guess automation, containerization, they are not, they're not exclusive to one another and they're not dependent on one another, but they very much work in and enhance one another. If we break containerization down, first of all.

So if you take a, a network component, traditionally it was a physical device. So in our world, the session border controller would be a physical device that operators would install in racks in their, in their environment. That evolved to be virtualized. So that's the first step in virtualization, which is a VNF.

A virtual network function, but it was still the same software stack. Not running on a purposeful, a dedicated piece of hardware, but running on a, a COTS, uh, Dell server, HP server, something like that. But it was still dedicated to that task when we moved to containerization. We have much more flexibility to be able to break that component part down into individual parts and much more rapidly spin them up in.

In the virtual cloud environment. The best way to explain containerization, I think is the analogy I use, is if you think about, uh, building a house, you build a house, you build a structure, there's multiple rooms inside the house. There's the plumbing, the electrics, the drainage. All of that's built for one purpose.

And you live in that house. Compare that to a hotel. A hotel. The infrastructure's there, the plumbing's there, the electric's there, the drainage, the broadband's all there for the, the infrastructure. And then you go in and you rent a room. You rent a room and you stay in the room, so it's much faster for you to actually just get a room.

You don't have to build the hotel. All the infrastructure is there, all the services and connectivity is there. Yeah, so you just have a room in the hotel 

Tom Arbuthot: and, and, and I guess you can obviously have a room. You can have six rooms. You have 20 rooms. You have 50 rooms. You can have 50 rooms for one night and 10 rooms for 10 nights.

Like you've got the, the flexibility of. Scale there as well. Completely. 

Matt Hurst: Exactly that and you, you could scale up the number of rooms you have and you don't have to change the infrastructure, the underlying connectivity, it's all there. So it's not faster and containerization is exactly that. We can take like the session border control and we can break it up into modules and can sort of pods.

One does, maybe the signaling piece, one maybe does the transcoding piece. One does the management piece. You don't need to scale the entire thing to make it bigger and add more capacity. Maybe just transcoding you need more of, so you can add more of those pods very, very quickly in a container of. 

Tom Arbuthot: And, and that's a, that you picked transcoding, that's a classic use case.

Right? I remember when I was specking, we're going back now, specking, like actual infrastructures and like transcoding was a big one, right? Yeah. Like, oh, you, your peak concurrency is this, that people might be using this code. It we're going between these two PBXs or PBX and Skype as it was. Yeah. Like we're gonna have to buy the right.

Physical infrastructure to deal with the peak scenario and rack it all and be ready. Yeah. And, and that would sometimes happen and sometimes it wouldn't. But there was a lot of slack in that process because you were like, you were building for the the peak concurrency, transcoding use case. 

Matt Hurst: Yeah, no, absolutely.

Right. And this is where now I can bring in some of the automation pieces. This is where the automation works for containerization to make that much more efficient so you don't have to. Overbuild the container environment, even for transcoding on SBC, you can build it to size to, to your appropriate needs to begin with, but then the automation piece comes on top to track the utilization, to track your, your traffic patterns and understand how your network operation runs over a long period of time.

And it understands that at certain times of the day, you have a, a peak of calls. You have he heavy transcoding the automation. Can automatically add transcoding pods in the container environment to meet that demand. And then when that demand fades away, the containers can be spun down. So you're much more efficient with, you know, your, your infrastructure and your resources that you are using.

Tom Arbuthot: That's so cool. And we would talk, like when we talk about this, I, I imagine a lot of people's immediately jump to public cloud. 'cause that's like the. Some of the driver for this has been like these massive infrastructures and we talked about hotels and renting. Yes, but, but this can be your own private infrastructure, either as a telco or as a larger enterprise.

You might already have some infrastructure you are going to couse for these scenarios. Yeah, 

Matt Hurst: absolutely. It doesn't have to be public cloud. You know, there are structures and the great thing with the containerization is you mentioned, you know, Kubernetes earlier. That's sort of an industry standard, uh, architecture that is in the public cloud, and it's also available in a private cloud environment as well.

Um, the containers are, you know, you can mix and match and you can use them in any of those environments. So yes, absolutely. A lot of telcos and certain enterprises and certain verticals have the need to have it in a private environment, and you can absolutely do that as well. And you can. 

Tom Arbuthot: Yeah. And the flexibility, I guess, like, like we might, we might choose to go.

Public now because it, it's faster, it's quicker, it's more cost efficient, whatever it may be. But we're not locking ourselves into that journey. Or actually, in some countries we might go public cloud. In some countries we might have our own infrastructure. Yeah. But architecturally we're using the same kind of approach.

Matt Hurst: Yeah, absolutely. Using the same approach. You're using the same. Industry standard tools. So, um, it fits very much into what we call the DevOps, um, environment. So DevOps with pipelines and, you know, Git repositories and everything else. This is very much an industry standard term, not just in the telco world and not just for our infrastructure components.

For, for everything. So. You know, the architecture that we're building for all of our containerized applications fits very much in that. So you can build up a, a DevOps pipeline to automate the deployment of, you know, multiple network components, you know, routers, switches, and then the SBCs management systems, routing platforms.

All of those can be fully put in the same architecture. One of the other great benefits of this, this kind of containerizing, the architect, the DevOps pipeline is single source of truth. So the, if you think about, as you said earlier, what the way you used a traditionally deployed network device or an SBC, even if it was a virtual device, you would configure that manually for that site, customize it.

Um, probably take a backup of the, that configuration, but that would just run in the, in the network and service that customer, if that device was to, to die and to fail. You'd have to restore the backup, maybe make some tweaks to that, to restore that function with the containerization and all that automation, you have a single source of truth, so you make the configuration in like the Git environment.

The actual source of truth for all your network components. The network components basically replicate themselves to that when they, they automatically spin up. If they die, the system automatically spawns a new container and it automatically takes that configuration, single source of truth. So you haven't gotta worry about configuring site by site.

It's all done from a central repository.. 

Tom Arbuthot: Yeah, you might have heard the, uh, kind of, I think infrastructure as code was a term Yeah. Banded around, but like, like, like, like real, real truth in like, this is our configuration and it is mapped to the actual infrastructure, not, I, I manually documented the infrastructure, or I took it back up six months ago.

Like, it's like this is the source of truth. 

Matt Hurst: Yeah, exactly. Right. And then you can do, within the automation stack, you can do periodic checks of your real network components to make sure they're still matching that source of truth. Or something changed, and then you can highlight that, oh, someone's manually made a change to a device, you know, inspect that or, or, or automatically overwrite that, maybe back to how it should be.

So it, it gives a, a much easier way for operators, for, for enterprises to make sure their network is running exactly how they think it should be, and it is configured. 

Tom Arbuthot: And Ribbon are really leading this in the, the telco space was, there are some big competitors there and, and a lot of people listening to our podcast will know Ribbon from the biggest tier one telcos.

Yeah. You know, into enterprise. Um, but this is a, uh, a bigger industry approach that you are leveraging here. Like software has been doing this for. Longer and, and containerization and gear and, and all this stuff. You are leveraging kind of years of experience in the, in the more software infrastructure realm to, to take to network and hardware and, and to not just at voice, but network as well.

Matt Hurst: Yeah, absolutely. It is full network, so, you know, as you know, Ribbon also do the sort of, the packet optical, um, and the IP routing stack as well. So it's taken it across the entire stack. Um, but yeah, we, I mean we started our virtualization journey, if you like, of our traditional, you know, pieces of tin hardware many, many years ago, um, you know, in nine, 10 years ago.

So we virtualized the, the first piece was to virtualize, so it can go on a VNF on a COTS, uh, server. That was done many years ago, and we spent the last sort of, uh, seven or eight years. Really perfecting the containerization and that optimization of resources. 'cause you know, as our discussion on transcoding early, we have so many requirements from customers where they need that ability to auto flex up and flex down in a public cloud or a private cloud environment.

So, you know, we've been doing that for a number of years. We've got customers who are sort of live with that. Um, it is kind of the, the next step for us is to overlay that intelligence and that automation and the AI layer on top to, to take that journey further. And it is a journey for telcos. Um, so there is a sort of industry standard body, uh, the TM forum, tele-management forum.

And they've actually defined sort of seven layers of automation from a fully manual process network to a fully autonomous network. And all of our customers are somewhere along that journey. Uh, some are further forward than others, but it's real. It really is happening now. Um, you know, we are working with some tier one operators in Europe who are.

Fully committed to making their network completely autonomous. It's, it's not about removing the humans and having the, the system run itself, but it's about improving the way that they can upgrade, automating the way they do testing, for example. Mm-hmm. So their, their operations team can be used for higher value tasks and creating new revenue rather than spending time on mundane, you know, upgrades and testing.

Tom Arbuthot: Yeah, like manual patching. Exactly. Yeah. 

Matt Hurst: Yeah. 

Tom Arbuthot: Yeah. And, and you mentioned it at the start of the conversation, but telco's on that journey, the automation conversation and the containerization conversation, they're obviously related and, and containerization lets you do cleverer things with automatization, which lets the AI do cleverer things.

Yeah. But there's, there's a bunch of automation you can do across the kind of asset quote unquote infrastructure as well. Yeah, 

Matt Hurst: exactly. Yeah, that's a very good point. You know, that, like I say, containerization and, and the. The automation go hand in hand and they compliment each other, but it's, they're not exclusive.

So, uh, yes, we have customers that have, you know, physical hardware devices, virtual devices, um, from Ribbon, and they are using our, um, our acumen automation platform on top to help with like test automation, upgrade automation that can still be done. Um, obviously the bit that can't be done is auto spinning up pods and things.

That's, that's where the containers. 

Tom Arbuthot: The, the automation can't quite rack and stack 

Matt Hurst: that. No, exactly. Yeah. That's a bit of a challenge. Yeah. I think the AI's got a long way to go before it's able to do that. 

Tom Arbuthot: So, uh, like, let's, let's zero in on telco because that, that's really interesting. Obviously that's our, our, my traditional space and, and Ribbon's sort of Strong Point, like.

How do telcos start that journey or, or to ask it a different way? What are the business pressures that are saying, oh, we're doing something new or different? Like, let's now look at. Containerization for this scenario or automation across our systems? 

Matt Hurst: Yeah. So the engagements we're having with customers and, and, and every customer is slightly different, but the con the conversation is always starts with what's the business problem?

What are we trying to achieve with you as a customer? Um, what, what are your pain points or what are your objectives for the next, uh, year, three years, five years? What are you trying to achieve? Are you trying to build more revenue? Are you trying to expand into new services? Um, are you trying to be more efficient and compete, or do you have a visibility problem across your infrastructure?

What you know that we can help fix? So very much the engagement approach for this automation journey is very much consultative one. So we have a dedicated team who work alongside our, our, uh, sales and system engineers. To work with customers to understand what those objectives are, and then build a together, build a proposal around what, what automation, where do we start?

We're not gonna do a big bang and everything on day one, but let's start with some of the smaller objectives we wanna achieve. Prove that out. Prove the ROI on that for the customer, obviously. Um, 'cause obviously the money, ma, the money piece matters. Yeah, of course. Um, and then grow from there. And we are seeing our customers, once they can start to realize that first step in automation, they realize the power and what it brings for 'em, and then they start to ramp up and scale up from there.

Telco, as you pointed out, telco is a different world because there are a lot of regulations country by country. Um, there's a lot of local laws, you know, lawful intercepts. There's a lot of liabilities on telcos that perhaps an enterprise doesn't have. So that's traditionally what makes them maybe a little bit slower to adopt some of this new technology.

Tom Arbuthot: And I, I've certainly seen myself, it seems like, uh, breaking into a new site or a new location. Like there's, there's increasing conversations about, uh, where, where infrastructure is, and particularly in Europe and like the ability to be like, okay, we're gonna spin up something new in, yeah. France, Germany, Spain, and we want to have it in country.

In country on a public cloud is a much faster starting point than in country. We're gonna take some colo or in country we're gonna have our own DC Yeah. Um, so that seems to be a jump off point for somebody to be like, okay, we've got our traditional infrastructure for, for, for this new site. We're not sure if we're gonna do a thousand customers or 10,000 customers yet.

We want flexibility, we want agility. We wanna get there in, in months, not years. Or maybe even over a year, not three years. Yeah. Which is, is very realistic in telco. Um, uh, have you seen that too? Is that kind of a, an interesting jump off point for companies to go from classic infrastructure to containerized?

Yeah. 

Matt Hurst: You know, where you've got telcos that are breaking out to new regions, new territories. Absolutely. Right. They're not, they're now not deploying a box into a data center. They're like, how can we quickly get that up and running, prove the market out? As you say, they're not quite sure. They've done their research, they've done their market analysis.

They've had a couple of marquee customers maybe, but they don't know how big it's gonna grow. So they will start with, you know, the virtualization suits them perfectly in a containerization so they can just spin up in a public cloud. Um, see how the business grows, see how, what scale it goes to, and then later on if they want to then move that to a private instance, um, they can do that.

So. You're right. That's where we're seeing the them start. Um, because it's, they don't have to do a massive forklift change in their infrastructure because the more traditional hardware platforms, the virtualized platforms, the containerized platforms, they can all work together. It's the same function.

It's just the underlying, how it's built. It's different. 

Tom Arbuthot: Awesome. And we've mostly talked about telco in on, in the pod, but what kind of enterprise, what scale of enterprise is, is. Doing this or what are the drivers for enterprise to do it? 

Matt Hurst: It is lar larger enterprise. Um, the main drivers are around sort of the operations piece, how they can be more efficient.

And we talked about, you know, SecOps, the security operations, how can they keep up to date with patching. Um, if you look in the, you know, the, the FinTech finance type of areas, there's a, a massive pressure and demand obviously to keep their security top notch. Um, and they don't have the resources to be, you know, checking vulnerabilities, you know, working with vendors to understand are there any third party components that are vulnerable and need patching.

So they are very much around uh, the conversation on automating that piece and all. 

Tom Arbuthot: Interesting. So it's not so much automating about flexibility of scale 'cause they're not maybe breaking into a new country and not knowing how many users they're gonna have. It's about that. Security. Visibility, operations.

Matt Hurst: Yeah, exactly. And also test. Test automation as well. So if they want to take a new software load or their, you know, their automation, they're gonna upgrade certain network components, how can they, you know, be sure that it's not gonna break anything. So they use our test automation tool where it can create a sort of a digital twin of the network in a lab environment.

Do the software upgrade. Run some automated tests that actually represent their call flows. So what we do is we actually do learning tra traffic pattern learning across their network over an extended period of time. And we break it down to, right here's your 500 most typical call flows, you know, refers, you know, internal, external, uh, call hold, that type of thing.

And we generate test scripts that you can run automatically against a new software load, make sure everything's running. Before you then roll it out to production. So a lot of large enterprise customers are, are really looking at that as well. 

Tom Arbuthot: And that's really interesting. Uh, the thing I've seen in enterprise is that that skillset of, of classic SBC management is getting thinner on the ground as well.

So actually having an operationalized system that is multiple people understand and that you've not got this, I've seen it where you've got two people who know how the SBC infrastructure works and, and that's about it. It's kind of a bit of black hole. Yeah. And I've just, just recorded a podcast on SOC 2, type 2 and some of the certifications that particularly hit, uh, finance.

And, and you need to have not just documented processes, but proven processes for your, your version management, your patch management, your maintaining infrastructure, your audit logs, all that stuff, uh, uh, has to be spot on to get through that. And then traditionally, the. I, I know because I worked at SBC traditionally, that has been an a, like a, a, a, a challenging area because it wasn't managed by the, the Wintel team patching the servers or the Linux team patching the servers.

It was kind of on its own island. Yeah. 

Matt Hurst: Correct. And, and this is where, you know, the, you mentioned the lack of skillset becomes a problem as well. You know, the operations teams are having to do more with less people and more with perhaps people that aren't super skilled in a particular product or a particular technology.

And that's where another piece of like the AI comes in is, you know, we, we give the ability, you can build a, like a chat bot. They can interact with and that chat bot can ingest all of the, obviously the, the Ribbon product documentation, but also the domain knowledge for the customer. Um, and then it can interface with trouble ticketing systems, but it can give like an interface to a junior engineer that they can interact with.

I'm seeing this error code 504, come on, this particular SBC. What does that mean? Where should I look next? They can interact with the chat bot that's got all that wealth of knowledge in the backend. So the junior engineer doesn't have to be an expert on a Ribbon SBC, uh, they don't have to be a VoIP expert.

They can be guided through with that sort of AI system. 

Tom Arbuthot: That's awesome. It's so cool to see this LLM technology built being pulled to a line of business use case, like you're saying. The interesting thing there is obviously you could put an LLM over public documentation. Yeah. That's one thing. But then pulling in the, the texting system and the history and being able to, you know, kind of, it, it, you can see where this is going.

Like it's going to start to have. Understanding of your infrastructure as well. Right. So actually you can see a future point where it's like, I've seen the, the the 500 and I recommend this, this, and this. 

Matt Hurst: Yeah, exactly. And it's that full stack visibility really comes into play. So we look across the, you know, the whole ISO stack.

So, and as you, as you know, it's, it is like Domino's, so you stack it up. The application at the top might be seeing a, a problem with, you know, core quality, for example. But that is, is that an underlying network problem? Is there congestion has a, has a router failed. Anything that happens at the lower levels impacts everything above it.

And by having sort of the AI and the analytics being able to look across the whole stack, they're then able to guide, uh, the, the technicians and the operations team to say, right, most probable cause for this particular error. They can look down the whole stack of saying, well, we've just seen this particular router has an outage or a high memory utilization.

You need to go and look there. 

Tom Arbuthot: It's so exciting to see that level of like, uh, technology being made available to everyone. Not just like, like obviously some teams have got amazing engineers that actually have the time and capacity to a look over the infrastructure and look at the logs and understand them.

But even with the best team in the world, you can't match the ability of these LMS to be on it with the data, the insights, the documentation, the level of. Like data they can consume 24/7. It's just a different gear. It it 

Matt Hurst: is. And you know, the, the LLM is there 24 hours a day. It's there at 2:00 AM rather than you having to wake up that senior engineer and wake him to have a coffee before he can function and help you with that question.

So yeah, it is phenomenal. It, it really is. And I think as more and more these real use cases with AI is gonna cut through some of the hype of AI and customers are actually really seeing the benefit. And, and the power of how it can improve their customer service. It can proactively, you know, head off a problem before it becomes a critical problem, for example.

Tom Arbuthot: Amazing. And Matt, thanks for taking you through that. Really interesting to hear how the containerization, the automation, and the AI story all come together. And some of the jump off points for that. If people wanna find out more or have a more depth conversation about their infrastructure and their requirements, what's the best thing to do?

Matt Hurst: Yeah, I mean, have a look on our, uh, Ribbon Communications homepage. Uh, obviously contact your local, uh, system engineers or sales team. They're more than happy to have a conversation. 

Tom Arbuthot: Awesome. 

Matt Hurst: Thanks so much, Matt. Thanks Tom. Appreciate it.