Moving Towards 100% Network Availability at Swisscom
Swisscom is the leading provider of communication, IT, and entertainment in Switzerland. Our services include TV, broadband, mobile, cloud, VOIP, and more—whether it’s for residential or the business market, we provide it. Like many service providers, we’ve evolved our service offerings to meet the needs and requirements of our clients.
These evolutions have come quickly, though, resulting in Swisscom operating a massively intricate system of distinct networks, each specified to the requirements of their respective service.
I have been building service provider networks in various capacities for a long time. I started my career building one of the first MPLS commercial networks, and even though I’ve ventured into IT territory, my true love has always been network design. Today, I’m the Lead Architect of Networks and Infrastructure at Swisscom, and I’ve seen firsthand how our challenges compounded as our network developed over the decades.
A Complex Network with Complex Challenges
Historically, we developed distinct IP networks for each of our services: an IP network for mobile, another for B2B services, yet another for IT services, and so on. We combined everything into service bundles, which also meant interconnecting all these networks, adding to the complexity. Eventually, it became one conglomeration of various networks. A little over a year ago, we performed a network analysis and realized our network infrastructure was getting too complex.
One problem is that if you make a change in one part of your network, it can influence a completely other part. When something went wrong, our teams made changes to resolve the issue but didn’t necessarily understand the impact it had on other areas and services.
Also, when working with multiple vendors, questions of responsibility arise. If something does not work well, accountability gets unclear. As the service provider, you’re the one stuck in the middle having to solve it.
In our team, we envisioned a completely new end-to-end design for just one network, rather than multiple interconnecting ones. We needed to remove complexity and converge services into a flexible and robust network that could satisfy the different requirements and Service Level Agreements (SLAs) for each service. Like all operators, we are under cost pressure, so part of our drive to decrease complexity was our desire to cut operating costs.
Finally, we needed a foundation where all current services could converge, with common protocols, security, and granular traffic control. That foundation would have to be standardized to be ready for anything and everything that 5G and other future applications could throw at us.
Looking to the Global Leader in Networking
Given its current prevalence, an IP-based network was the obvious choice. Next came choosing a partner. Swisscom had worked closely with Cisco for more than 25 years, following their network innovations in IP infrastructure closely, but we had also worked with a variety of other vendors. Not many have the full-service portfolio that Cisco has, which is necessary for an end-to-end design from the data centers to the central offices.
We wanted a partner with a lot of development power, and we see Cisco as a competitive, innovative network developer as well as a leader in IP networking. Many major IP-networking developments, from MPLS to segment routing, started with Cisco and later became world standards. That gave us the confidence that Cisco could help us create a more reliable, resilient network that would not only be state of the art today but that would incorporate future trends.
That expertise was important for a project like this, which is significantly different from a normal project that begins with an end in mind. Typically, when you tender a project, you pick the best solution among today’s technical offerings. In our case, we expected the transformation to take place over the course of five years. But over that time, new network components may come to market, and no vendor can give you an accurate cost or specification for a component you intend to order that far in the future. So we had to take a different approach to this project: We selected the partner we considered to be a market leader, and then worked with them to define the target architecture.
A Unique Architecture That Ensures the Best Possible Experience
When it comes to designing the architecture over such a long period, we started at a very high level. We set out a roadmap with goal posts for each year based in part on which solutions are nearing the end of their lifecycle, then we got into the technical details. This is where working with Cisco’s Customer Experience (CX) team proved invaluable. They brought in their specialists, product managers, and architecture team to explain to us what they saw coming down the pipeline, and combined this with our experience with the daily operations of IP networks and services. We know there’s more than one way to get to Rome, as the saying goes, and yet we want to benefit from the solutions Cisco sees coming. This is how we are going to achieve the best possible architecture in five years.
In collaboration between Swisscom architects and Cisco engineers, our aspiration for a single, simplified network has been realized through segment routing. As all network operators, we have always built with MPLS. You combine different MPLS networks and call that Seamless MPLS. But in doing so, you have multiple labels flowing through the networks, causing scalability and complexity issues. Segment routing is a much newer technology. It’s easier to operate and scale, and is more cost effective. Greater complexity also leads to a greater risk of incidents and issues, so a simpler network is also a more stable one.
Using segment routing, Swisscom will be able to offer guaranteed end-to-end SLAs for different services to our enterprise customers, ensuring latency-sensitive traffic is auto-routed down the shortest, least congested paths. We can even offer secure, encrypted sub-topologies over one network.
We can make that guarantee because we have designed the network to achieve our goal of 100% availability. On a single backbone, that goal is impossible. No matter how good your network or how much monitoring you have, one day you will have an outage. It might only be for a few minutes, and it might happen only once every few years, but it will happen. That seems insignificant, but for a provider with a large market share, it would have a huge impact on Swiss society.
That’s why we decided we would not build one core network, but two. In the rare occasion that a complete core network goes down, all traffic will then transfer to the other network. Segment routing supports this kind of architecture. This will also prepare us for emerging 5G applications and services, with the ability to integrate network-slicing policies to make sure each service has the best possible experience.
This will be a big differentiator for us in the B2B market, where reliability of the network is key and capacity is a secondary consideration. With two core networks and segment routing, Swisscom will guarantee complete, continual connectivity.
Automation and Monitoring for a Simpler Operation
In addition to granular traffic controls and offering end-to-end SLAs, having an IP-based network means we can reduce the operational complexity of the network with automation. I’m particularly proud of the telemetry-based monitoring we’ve incorporated into our new network, and that’s improving every month.
We are in the early days of our rollout and our environment is still rather complex, but we have a great team working on monitoring. We already have far better insights into the network, which allows us to solve future issues proactively. Using telemetry data from various network sources, our engineers can closely monitor the network and implement pre-planned responses automatically as traffic demands shift.
Automated responses and intelligent alerts mean our engineers will spend less time troubleshooting while minimizing large customer-impacting events. It can happen that one part of the network fails, but the root cause is somewhere else in the network entirely. In a complex network, locating that root cause can take hours. With better control over a simplified network, you can reduce that reaction time to minutes, or even seconds.
The Path to a Simpler Network Starts Here
After selecting Cisco as our partner at the beginning of 2021, we are now six months into the project and have implemented the mobile RAN. We want to modernize the IT to enable NetDevOps and CI/CD pipelines, because once you have CI/CD in place, it’s much easier to roll out the network. By 2026, we plan to have replaced at least 80% of our current infrastructure. It will allow Swisscom to expand services, scale with client demand, and significantly reduce costs in the process.
This convergence provides the ability to drastically reduce the amount of nodes required to run the network. We expect the potential power savings to be in the region of 40%, thus enabling a more sustainable platform for Switzerland’s digital economy.
Rather than relying on customized solutions, we're moving more toward more mainstream, standardized ones. Cisco has helped us begin to implement a state-of-the-art solution that will be the world’s most robust and flexible, and they’ve provided us with the expertise to predict future trends.