Episode #25 15:11 2024-07-05

#025 – Kubernetes for Humans Podcast with Ashan Senevirathne & Joel Studler (Swisscom)

Ashan Senevirathne
Product Owner, Mobile Cloud Native Automation & DevOps Engineer, Swisscom
Joel Studler
DevOps Engineer on the team building Swisscom's cloud-native 5G core.

Listen to the Podcast

Episode Overview

In this episode of Kubernetes for Humans, host Itiel Shwartz sits down with Ashan Senevirathne and Joel Studler from Swisscom to unpack what it really takes to run a 5G core on Kubernetes. They walk through the cultural and technical shift from a vendor-driven, VM-based 4G stack to a white-box, cloud-native 5G core, the operational overhead of layering Telco lifecycle tooling on top of Kubernetes, and the scale challenge of managing roughly 2,000 pods just for a small dev deployment across four data centers serving Switzerland's nine million users. The conversation digs into why blue-green and canary deployments break down at Telco scale, how the team is collapsing out-of-band orchestration into the Kubernetes API via operators and CRDs, and the industry standards work happening across CNCF, LF Networking, Nephio, and ETSI to make cloud-native Telco a reality.

In this episode we discuss:

  • What the 5G core actually is and why Telcos are moving it from VMs to containers on Kubernetes
  • The cultural shift from vendor-driven 4G integration to a white-box, software-mindset approach in 5G
  • Operational overhead of running out-of-band Telco lifecycle tooling alongside Kubernetes — and collapsing it into the Kubernetes API
  • Scale realities: ~2,000 pods in a small dev environment across four data centers, and why blue-green and canary don't trivially apply
  • Industry standards and communities driving cloud-native Telco: CNCF, LF Networking, Nephio, ETSI, and tools like Nokia's SDC

Key Takeaways

1
Adopting Kubernetes for the 5G core is a 2–3 year journey, gated as much by vendor readiness as by Kubernetes itself — many Telco vendors are still rewriting their stacks to fit cloud-native patterns.
2
Telco apps aren't ordinary cloud apps: configuration and lifecycle management are far more complex, so the real work is building the surrounding automation, not just running containers.
3
Layering separate orchestration tools on top of Kubernetes duplicates effort; extending the Kubernetes API with operators and CRDs lets reconciliation handle Telco lifecycle natively.
4
At ~2,000 pods per deployment, classic blue-green and canary strategies stop working — teams have to lean on in-service upgrades and modular per-component lifecycles.
5
The next wave of Kubernetes wins for Telco won't come from more features — they'll come from better user experience, stronger security, and energy-efficient, sustainable platforms.

Itiel Shwartz:  Hello everyone, and welcome to our live episode of the Kubernetes for Humans podcast. With me on the show today, we’re going to talk about the journey to Kubernetes, lessons learned, scars, and so on. Maybe we can start by having you talk a bit about yourselves, giving us some background on who you are, what you do, and how you got into Kubernetes.

Ashan Senevirathne:  Yeah, so my name is Ashan. I’m the Product Owner of the Mobile Cloud Native Automation team. What we do at Swisscom is we’re part of the mobile organization. Swisscom is a large company, and within our organization, we’re aiming to deploy the 5G core on Kubernetes because it comes as microservices in a cloud-native way. The main responsibility of our team is to implement this framework so it’s on the right path to being cloud-native.

Joel Studler:  I’m a DevOps engineer. I’ve been in platform engineering for a while now. I have a big infrastructure background, and then I moved more into the cloud-native Kubernetes area. I have a broad background in platform engineering, and now we’re really moving towards Kubernetes operators and automating as much as possible in the cluster.

Itiel Shwartz:  Usually, when people think about 5G, they think about edge computing or antennas. Can you share what 5G is and how it relates to Kubernetes? Why is Kubernetes needed for this?

Joel Studler:  When we talk about 5G, it’s mainly two areas. One area is the Radio Access Network, like the antennas — the part that makes your cell phone say “5G.” That’s being rolled out globally and is in progress. The other part is the core, the centralized deployment where all the magic happens, like user management, authorization, authentication, mapping your SIM card to a profile, and so on. It’s also responsible for forwarding data and voice calls. We’re seeing a trend of moving from VMs, which the 4G stack was built on, to containers in 5G. That’s the evolution. A while back, like 20 to 25 years ago, we had bare metal or custom hardware.

Itiel Shwartz:  So is this something next to the antenna, or is it on some cloud or data center?

Ashan Senevirathne:  In terms of the antennas, we have thousands of them. For the core, there are only about four main data centers. The antennas provide the wireless capability, and the core handles the logic and smartness within the network. The transition from 4G to 5G involves moving from VMs to containers. What’s also important is the culture within the company because operating a 4G network is quite different from a 5G network.

Itiel Shwartz:  Why is it a cultural shift to move from 4G to 5G?

Ashan Senevirathne:  The shift is about having more of a software mindset within the company. In our organization, we’re pretty much vendor-driven. We don’t write the software applications for the network functions like the 5G core; we’re more like software integrators. In the 4G and 3G worlds, we were just integrating, but now we’re taking more of a white-box approach, doing the integration ourselves and implementing the surrounding framework for automation.

Itiel Shwartz:  How long did it take you to adopt Kubernetes? I guess telecom generally moves slowly, but maybe I’m wrong.

Joel Studler:  For the 5G core, it’s taken us about two to three years, and we’re still ongoing. We haven’t solved all the problems. It’s a mix of vendors not being ready and needing to rebuild many of their software projects. They’re just not ready yet. Another challenge is building the entire automation around it because Telco apps are much more complex than the average cloud app. The industry doesn’t have a clear plan on how to solve these issues, so we’re trying to push forward and innovate, hoping others can benefit eventually.

Ashan Senevirathne:  Yes, and with Telco apps, you have to consider the lifecycle aspect and all the changes that need to be made to the workloads. The big challenge is how to make it more cloud-native because there are a lot of out-of-band Kubernetes operations that need to be brought into the Kubernetes layer.

Itiel Shwartz:  What’s the scale of this deployment?

Joel Studler:  We’re not live yet with the 5G core, but we plan to go live towards the end of the year or early next year. The scale involves about four data centers with roughly 80 blades per site. In Switzerland, we’re talking about millions of users — roughly nine million citizens. Everything needs to be very low latency and fault-tolerant.

Ashan Senevirathne:  Switzerland probably has one of the best networks globally in terms of reliability and stability during calls and handovers between antennas.

Joel Studler:  Latency isn’t much of a problem in Switzerland because the country is small, so we’re only talking about one or two milliseconds that could be gained, which would be expensive to achieve by building more data centers.

Itiel Shwartz:  What was the biggest hurdle or mistake you had to overcome in this adoption process?

Ashan Senevirathne:  I’d say there are two main aspects. The first is the lifecycle of the core or the network. We found that Kubernetes was doing the orchestration, but we also had another set of tools doing the same thing, which added a lot of operational overhead. Now, we’re working on bringing everything to the Kubernetes layer by extending the Kubernetes API. The second aspect is exploring public cloud options. The application needs to be cloud-native to fully benefit from hyperscalers.

Joel Studler:  Another important aspect is the scale of the application. We’re talking about 2,000 pods in a dev environment for a small deployment. This makes it difficult to do blue-green deployments or canary deployments. The software doesn’t support it, and managing 2,000 containers is a significant challenge. We’re still figuring it out, but one idea is to enable in-service upgrades where the pods survive, and you just change the configuration or parameters. Another idea is to modularize the application so that different pieces can have their own lifecycle.

Itiel Shwartz:  Is there any collaboration or standards in place for adopting 5G with Kubernetes globally?

Ashan Senevirathne:  There’s a working group within the Linux Foundation and also within CNCF. Many operators, including Swisscom, are involved. Last year, we published a white paper called “Transformation Journey to Cloud Native Applications,” which covers the requirements for cloud-native network functions. There are also other standards like the European Telecommunications Standards Institute (ETSI).

Joel Studler:  There are communities as well, like the Linux Foundation Networking group, which brings out projects like Nephio that tackle these topics. Nokia and others are pushing forward on these fronts. We actually demoed a tool called SDC this afternoon. It’s a new tool for config management for Telco workloads, initially developed by Nokia.

Itiel Shwartz:  Any predictions for the future of Kubernetes or the space in general? Where do you think the industry will be in two years?

Joel Studler:  What I hope for is that we’ll work on improving user experience and simplicity. Kubernetes is still complex, and you need to know a lot to manage it. The developer experience isn’t that great compared to other tools, so I hope we see improvements in that area.

Ashan Senevirathne:  I agree with Joel about simplicity. I’d also like to see more focus on security and how we can build reliable and sustainable applications. There’s a big push from Swisscom for energy efficiency and green energy, so I’d like to see those aspects being addressed as well.

Itiel Shwartz:  That’s super cool. I think we’ve covered most of the important topics. I know you guys have a talk coming up, but I’m not sure if people will be able to attend. There should be a recording available, though, right?

Ashan Senevirathne:  Yes, there will be a recording, and we’ll try to have a YouTube link for that.

Itiel Shwartz:  Thanks very much, and enjoy the rest of your conference!

Ashan Senevirathne and Joel Studler:  Thank you!

This is an AI generated transcript of the conversation

About the Guest

Ashan Senevirathne
Product Owner, Mobile Cloud Native Automation & DevOps Engineer, Swisscom
Ashan Senevirathne is an experienced Product Owner and Senior DevOps Engineer with a proven track record in driving innovation and efficiency in telecommunications. At Swisscom, he leads the development of a cloud-native orchestration framework for 5G Core using Kubernetes. He is recognized for his expertise in Kubernetes, GitOps, cloud-native principles, CI/CD, and network orchestration.
Joel Studler
DevOps Engineer on the team building Swisscom's cloud-native 5G core.
Joel Studler is a DevOps Engineer on the team building Swisscom's cloud-native 5G core. He brings a deep background in infrastructure automation, software-defined networking, and highly available databases, is CK* certified, and has authored several CRD/Operator-based Kubernetes extensions. Joel is passionate about automation and developer experience; outside tech, he's an active musician on stage and in the studio.