There is no “silver bullet” for moving network workloads to the cloud, according to Telefónica Germany CTIO, speaking at this year's Mobile World Congress.
Telefónica Germany CTIO outlines network cloud challenges
Telefónica’s vision for the “Network of the Future” is to disaggregate every component and run them in the cloud. Its aim is to build a foundation for new revenue models and growth by enabling network capabilities to be exposed through open interfaces to a bigger community of developers and potential new users. At Mobile World Congress the operator shared insights on how it is adapting to the cloud in its network and IT transformation.
The full future network vision has four components: retiring legacy networks, artificial intelligence (AI) and automation, open architectures and going to the cloud.
Cloudifying the network, however, is a lot harder than moving telcos IT applications to the cloud, according to the telco’s German operating business, which has been working towards this ambition for the last three years.
Telefónica Germany has already moved a bulk of IT applications into the public cloud as part of a five-year, €600 million Radical Architecture IT Transformation program. It aims to have 80% of IT workloads sitting in the public cloud and, by the end of this year, “€1 billion revenues flowing through the platform we created,” said Mallik Rao, Chief Technology & Information Officer at Telefónica Germany, speaking during a presentation at MWC.
To give an idea of how much things have changed in three years on the IT side, Rao said the operator used to spend around €1.5 million to €2 million per year with hyperscalers and today that annual spend is nearly €35 million.
Moving IT workloads to the cloud is “easy,” he said. Network workloads are more demanding, however, and Telefónica’s strategy to “cloudify” and “softwarize” the network means implementing workloads in a mix of private and multi-cloud environments that meet telco requirements.
“To take workloads into the cloud, there is no single silver bullet. We're not trying to hedge one against the other,” he said, suggesting some ecosystems are not able to handle the network functions.
“It's a misnomer to call it a private cloud and public cloud. Cloud is a technology. Where to host this technology is something which we are figuring out,” he added.
Telefónica’s broader ambition encompasses using cloud native features like CI/CD automated testing, auto-scaling and auto-healing to reduce time-to-market for services and improve reliability, as well as microservices-based network functions, end-to-end orchestration, and multi-cloud support.
“We need to evaluate what is the best cloud infrastructure … to deploy our network functions… We need to take into account the technical capabilities, the possibilities of automation, the regulatory requirements, and the economic conditions,” said Cayetano Carbajo, Telefónica’s Director of Core, Transport and Platforms for Global CTIO, speaking alongside Rao at MWC.
“There's no point in bringing a network function to the public cloud if I'm not able to increase my level of automation or increase my level of reliability. These are key things. We [also] need to fulfil our [total cost of ownership] requirements,” he added.
Telefónica Germany has explored public clouds for intensive network functions like the 5G core. Back in 2020, the telco worked with Ericsson and Amazon Web Services on to deploy 5G core in the cloud. The operator has also teamed with Google Cloud on 5G core implementations.
“We learned a lot [about] the operating model and the ability of the infrastructure to take a workload [that is] significantly transaction intensive. It is not an IT application. It is a real-time workload,” he said.
Rao said the wider industry needs to “come together and solve these problems,” because they cannot be resolved by operators on their own. For its part, Telefónica’s transformation is a group-wide effort. Much of the work is centered in Germany, and the learnings are shared with other operating businesses. Also, there are similar tests and trials going on in Brazil and the UK. Furthermore, Telefónica has “horizontal” programs throughout the group, such as for autonomous networks.
“We have learned in the last three years that is it is easy to take an application to the cloud, but it's extremely hard to manage it the way it should be managed in cloud. You can't repeat the way of managing in the on-premise world into the cloud,” said Rao, explaining that this will be a learning journey for the next five to 10 years.
Here, one of Telefónica’s top cloud criteria is the ability to auto-scale across an entire telco domain and not just in each application. For example, Rao said Telefónica Germany’s retail domain had about 14 different applications serving the operator’s shops and call centers. Another domain would be voice services.
This functionality is not there yet. But he said he hoped that by 2025, the operator could start demonstrating “at least two or three big domains in our own infrastructure.”
Rao described the journey of taking network functions into the cloud as a way of “taking back control” of the network. Operators have outsourced so much over the years that they have become “high-quality coordinators” of their suppliers.
“When you talk about disaggregation, it's a discipline that you want to have in your organisation. It's not about AWS or Google or Microsoft or private cloud. It's about us,” he said.
“We are spending an enormous amount of time in relearning the engineering of [the network]. Because if you go into cloud, if you don't understand the engineering of it, you can't operate,” he added.
Artificial intelligence (AI) and automation play key roles in Telefónica’s future network strategy. The operator has committed to achieving TM Forum’s Level 4 Autonomous Networks (AN) by 2025, one of few operators to do so along with AIS, China Mobile, MTN and Orange. It has also endorsed TM Forum’s AN Manifesto.
Carbajo said the operator is not waiting until 2025. It is already using a Large Language Model to have a “natural language interface” with some network nodes to ask, for example, which base stations have more traffic than others at specific times and locations. It is also working on automated root cause analysis.