Meet SARA, the AI That May Be a Big Deal for Austria's Main Telco

LONDON -- Telco AI Summit Europe -- Austrian telecom incumbent A1 has built its own AI-based platform to support the rollout and management of mobile networks and is now looking into the possibility of selling it to other operators.

Branded SARA, for Superior Analytics of RAN (radio access networks) by A1, the technology is designed to ease network planning and take some of the guesswork out of decisions about network upgrades and maintenance. It is already operational in Austria, North Macedonia, Serbia and Slovenia and is due to be launched soon in Bulgaria and Croatia.

Shown off at this week's Telco AI Summit in London, SARA has been in the works since 2016, when A1 -- part of the A1 Telekom Austria Group that is majority owned by Latin American telecom giant América Móvil -- grew concerned that take-up of a just-launched mobile broadband service might affect network quality.

Thomas Hodi, a software architect who worked on the project, said it could prove valuable given estimates by the GSM Association, an industry lobby group, that European operators will spend about $500 billion on the rollout of 5G networks in coming years.

"This large investment should be based not on gut feeling but on analytical decisions," he said during a presentation of the platform in London.

A1 developed SARA using a team of RAN experts from more than seven countries. For expertise in machine learning, a subset of AI that relies heavily on data analytics, it turned to a startup called Things Solver.

The final product has several applications, according to Hodi, and was first used to prevent A1 from overselling a 4G-based broadband service (so-called fixed wireless access) and encountering network problems. It does that monitoring and forecasting traffic at specific cells and then providing insights to its sales department.

Potentially even more important is what A1 calls "smart capex planning," which examines which sites are heavily congested and then calculates what levels of investment are justified and what the return on investment will be.

While the existing tool makes forecasts for the next three months, an update will allow the operator to plan for up to one year ahead and could prove vital when it comes to 5G investments. "We can see what the best antenna solution is based on coverage and future capacity needs," said Hodi.

SARA also appears to offer scope for long-term capacity planning, which would make forecasts about network requirements over a four-year period and help to answer strategic questions about spectrum re-farming with the shutdown of older mobile technologies.

France's Orange is working on a similar AI-based "smart capex" tool in Spain, although it is at a much earlier stage of development. "We are processing cell-level and customer-level data and using machine learning to predict capacity and coverage and to extend and optimize the networks," said Ludovic Lévy, Orange's group vice president of data strategy, here in London this week. He reckons the system could lead to a "two-digit-percentage" decrease in cell-level capex. "That is cash flow for the company.

Want to know more about 5G? Check out our dedicated 5G content channel here on
Light Reading.

What makes the A1 system especially intriguing is that it could also be marketed to other operators in future. A1 is about to start deploying SARA with Claro Colombia, which is also owned by América Móvil but which lies outside the A1 group. Discussions are underway about going further.

"Our main priority is internal, but we see that we have developed a platform that can scale" and which could be developed as a commercial product, Hodi told Light Reading when asked if A1 would sell its technology to other operators. "A decision has yet to be made."

Other use cases described in London include "anomaly detection," which aims to reduce the time spent resolving problems in the radio network, as well as something Hodi calls the "CM Browser."

"It is the Google for network parameters," he explains. The feature would allow engineers to search for different network parameters across the company's own network, as well as others, using SARA and could in future support technical collaboration between operators.

Related posts:

— Iain Morris, International Editor, Light Reading

Be the first to post a comment regarding this story.
Sign In