5GSA – for now the innovation is all under the hood at Swisscom
Swisscom’s Mark Düsener talks to Contributing Editor Annie Turner about the critical power of cloud native and automation – and why preparing properly for 5GSA right is more important than launching first.
Mark Düsener, Executive Vice President Mobile Network & Services at Swisscom, is a relative newcomer, having joined the operator in November 2020. Previously was at Deutsche Telekom’s T-Systems International in different roles over several years. He is a great example of why fresh thinking, even in a culture of excellence, can be such a powerful thing.
Swisscom is justifiably proud of being the first in Europe to launch 5G, then achieving 99% coverage of the population in less than six months. It has also been recognised many times for its world-beating infrastructure, including as having the country’s best 5G network by Ookla in terms of both speed and coverage.
Yet when it came to 5G Standalone (5GSA), he resisted his team’s “old habit” of being first because, “We don’t’ do things for the sake of doing them, but to enhance the quality and resilience of the network and the speed of innovation for our customers.” He adds, “There’s no customer demand yet and secondly, it’s a great opportunity to change the way we do things, because it’s a big technology shift.” In other words, getting it right is more important than being first – and that includes timing.
Customers come first
Düsener politely suggests that some announced 5GSA deployments are marketing gimmicks or proofs of concept more than commercial offers. He observes that popular showcases, such as broadcasting from stadia, “Don’t profoundly change an operator’s business.”
So where are new commercial use cases that will bring that change and what does he think they will be? “That’s different for each country. We’re a small country with very close to total coverage. Most European countries don’t have that coverage yet, so TV from a small rural stadium won’t work without 5G coverage, but that’s not holding us back in Switzerland. We only have four datacentres supporting our mobile infrastructure, so the distance between them isn’t great and we already have latencies of 10 milliseconds and guaranteed latency is a feature we can apply.”
Readiness is all
While various industrial sectors are readying themselves for the next stage of digitalization and automation, he is seizing the chance to be ready when they are. Düsener says, “I want to use this rare opportunity of a greenfield approach. It requires a different way of working. We’ve focused on setting up the optimisation chains, building the tools and training ourselves, exploring how to work differently.”
Swisscom did not embark on this journey talking about transformation, he explains, although it is mightily ambitious. He points out, “We are not only a telco, but a big IT business too – our logo says, ‘From telco to techco’. We do not only want to consume but to drive the softwarisation of our industry,” he states. Düsener’s starting point was building “people journeys” – looking at what new skills would be required for the cloud-native 5GSA architecture.
In particular, he was keen to embrace continuous integration, deployment and testing (CI/CD/CT), different forms of release management and optimisation programming. He says, “Only when we started that process of upskilling, which is up and running, did we start the transformation necessary to become a software organisation.”
Düsener stresses that cloud native is the driving force, adding, “It’s an industry effort and it is our ambition to lead the industry with our partners to achieve simpler, cheaper operations, greater optimization of network assets, better customer experience – all of these things.”
Start with service design
He elaborates, “It starts with service design. Instead of planning ‘this network function will get this IP address and so on’, we want an operating system, Kubernetes and the like, that automatically instantiates what’s necessary. We define our intent, what we need for connectivity, instead of describing how to do it, as we do today. The telecom model always about shifting and deploying physical boxes, then they became virtual, but are still treated as boxes. Now we have the chance to either have cloud-native boxes, or we get rid of the concept of boxes.
“Instead of thinking, ‘I need eight packet core gateways’, I don’t care how many I have. I need that core functionality, and the system will instantiate the number necessary – maybe adding a few in peak hours and shutting them down during the night.”
That’s the vision, what will it take to get there? Düsener says, “We need capabilities like connectivity as a service – I request connectivity and automation happens in the background network as a service that delivers some IP addresses and manages the resources and so on.”
Even further down the line, “I believe everything in the end will be code: I will be able to address functionality as code to decouple further. Today in core we typically have maybe one major release per year that we need to instantiate in the laboratory and test. It’s a really huge effort. I want releases per month or faster and not only ‘as required’ but ‘as delivered’. Once I deliver something – a patch or new product feature – I instantiate it with CD/CI and let’s not forget CT. That has a huge impact on how we work.”
Canaries and A/B testing
Swisscom has advanced furthest with this softwarisation is its radio access networks, which the most hardware intensive. The operator releases new features every second week. Düsener notes wryly, “I have heard that my peers would like to understand how we do it. We use ‘canary testing’ by applying new feature to a selected area for initial testing and even A/B testing where we try out two different feature sets on the same antenna with 50% of the customers using A and the others B, then we compare the results.”
The operator is optimising power usage for different frequency bands, but this depends on which kinds of phones are in use in any given area and the frequency bands they use. Power and other kinds of optimisation are automated, through more than 25 fully closed-loop use cases up that run constantly.
While Düsener expects the number of closed loops in use to proliferate, he does not see an estimate of the eventual number as a useful metric so much as keeping the automation vision in mind. In access, for example, the plan is to create a base-site configuration algorithm that the operator will use, unaltered, to configure all sites. Configuration and replication, not customisation, is the aim, although, as he says, “You’re never finished with optimisation.”
Being stateless
In the core, most of Swisscom’s systems that serve customers are virtualised with a few dedicated appliances still in use and a few cloud-native installations. “The next lever really would be 5GSA,” Düsener says, but in the meantime, the operator manages virtual packet and voice cores on its private telco cloud, but it is already working on “the next big step” with public cloud hyperscaler, AWS.
Düsener says, “I want to be able to deploy AWS not only as an outpost in my data centre, I want to be able to deploy real cloud native public cloud, not private cloud. Yes, this means a certain level of security and privacy, but more importantly, telco workloads aren’t cloud native today because they are session based, whereas being cloud native means you don’t have a state.
“When most workloads are built on cloud-native principles, the changes are manifest because most big workloads and most types of workloads use layer 2 networks and public clouds use layer 3”. He adds, “Now almost all our workloads are stateful, so changing is not an easy task, but if there is no state and a certain network function is not behaving properly, you just kill it. You don’t have to take care of anything else, the environment, Kubernetes and so on, will re-instantiate to meet the need automatically, based on the defined end state.”
Düsener continues, “It’s not an easy design choice, but it is a design choice. This is what we’re driving with our partners, AWS and Ericsson, because that’s what will change the industry. We believe there will be a big, dramatic, positive impact on how we operate.”
Automating innovation
When will this major change happen? Düsener is not to be drawn into a precise answer about when he thinks this will be, saying, “Less than five years more than six months…I think it’s a typical hockey stick scenario; you need you need certain factors to be in place before the wheel starts to turn faster. At the moment, we’re not driving 5GSA because we’re waiting for that increase in wheel speed.”
He stresses, “The real revolution is under the hood [or bonnet, if you’re British] It’s not about faster networks or the next use case. We will see some business-critical communications coming up and so on, and there will be use cases because of new features that we couldn’t do in the past. But once we are able to automate further, to really apply software patterns to our industry, the speed of innovation will change dramatically.”