Service providers see more network automation, AI from the core to the edge

Recent examples offered by Verizon and T-Mobile show a wide and growing variety of network automation, with some AI use, as 5G arrives.

Phil Harvey, Editor-in-Chief

December 24, 2020

5 Min Read
Light Reading logo in a gray background | Light Reading

Telecom networks are far from being sentient, but they are becoming more and more responsive and automated, which helps service providers lower their operating costs.

Two of the largest service providers in the US are using AI and machine learning in a variety of ways. Here's a quick glimpse of what they've talked about doing lately with improved network monitoring and testing and what's ahead as the network edge becomes more responsive to an influx of users, devices and changing network conditions.

Network monitoring

Adam Koeppe, SVP for technology strategy, architecture and planning at Verizon, said AI makes it possible to improve network performance by capturing data from the interactions of tens of thousands of network nodes with hundreds of millions of customers.

"The intelligence that comes out of all the componentry that's on the network is voluminous," Koeppe said. "Every one of those cell sites, every one of those transport connections, every one of those customer data interactions, generates something that we can use to better understand the quality of the experience."

T-Mobile is also using AI and machine learning for a whole portfolio of use cases involving network monitoring and management. Those include "anomaly detection in monitoring critical services, where we rely on third parties to deliver calls, such as 911 and messaging platforms, to closed-loop intelligence to predict radio network changes or temporary congestion and trigger automatic parameter changes," writes Brian King, T-Mobile's SVP and COO of technology, in an email to Light Reading.

Service providers are increasingly using AI in their radio access networks, as RAN elements become more distributed through small cells and virtualization. Verizon, for example, uses AI to enable self-organizing networks. When a site goes down, the operator uses AI and machine learning to remotely adjust antenna downtilt at other sites in order to compensate. "That happens real-time based on the numerous feeds that come out of the network," Koeppe explained.

Testing new features

T-Mobile said it uses automation and AI to test and validate network features and software upgrades before they go from the lab to the real world. "Within its lab environments, T-Mobile has an extensive array of automation tools that are able to simulate real-world telecommunications conditions, so that we can validate the features and functionality of handsets and new software and hardware that will be deployed in the T-Mobile 5G network," King said. "This testing includes things like automatic drop testing to ensure that new handsets are not overly susceptible to damage when accidentally dropped."

At the (cutting) edge

Network operators are excited about the ways that AI will help them leverage the capabilities of next-generation networks. But they'll need to resist the temptation to do everything at once. "The edge is uncharted waters for telcos," Stephanie Gibbons, principal analyst for carrier network software at Omdia, a sister company of Light Reading, said earlier this month.

Network slicing will help mobile network operators, allowing them to run multiple virtual networks within a single physical network. This goes hand-in-hand with edge computing, and both technologies will be made more robust by AI and ML. "These AI, ML and automation protocols feed right into how you end up slicing a fully virtualized network," said Verizon's Koeppe.

Verizon plans to use AI at the edge of the network to detect applications that need network resources and provide those resources in real time. Koeppe used the example of mobile gaming to explain. "The network is going to know that that type of device needs a certain throughput and a certain latency treatment in order to be an enjoyable experience for the customer," he said. "We'll be using AI and ML ... to ensure that that happens." This capability relies on both edge computing and network slicing, both of which can leverage AI.

As reported in November, Verizon sees its edge computing capabilities as a key enabler of new services and it expects to make money from them in the near future. The carrier has begun outlining the companies currently testing services powered by edge computing. Connected car startup Renovo is among those testing services on Verizon's edge network.

And, as T-Mobile's King notes, IoT devices and applications are also going to be requesting network resources and AI can shine here, too. "As the 5G network continues to expand into the virtualized and cloud native environments, AI will play a large role in monitoring and adjusting the various network components that may be used by customers or IoT devices and applications," he said.

Telecom investment in AI gets real

Verizon and T-Mobile aren't alone, of course. Communications service providers all over the world are hitting a point where the influx of new traffic and devices is more than legacy systems and manual processes can handle. Analysts are looking at AI and machine learning as key investment areas next year, as service providers prioritize their investments around saving opex and unlocking services on their new 5G networks.

"Automating the increasingly complex network and service management environment has become a major priority for CSPs, along with the AI and data management capabilities required to support this," writes Omdia Practice Leader Kris Szaniawski, in his report, "2021 Trends to Watch: Telecom Operations and IT."

According to Omdia's enterprise ICT surveys, nearly 80% of CSPs see the use of AI/analytics to automate network activities as an "important" or "very important" IT project for 2021. Nearly 60% of the CSPs that Omdia surveyed are planning to increase investment in AI tools.

Some of the top use cases for service provider AI include "network fault prediction and prevention, automation of end-to-end life-cycle management, and the management of network slicing," Szaniawski writes.

— Phil Harvey, Editor-in-Chief, Light Reading

Light Reading contributor Martha DeGrasse provided additional reporting for this story.

About the Author

Phil Harvey

Editor-in-Chief, Light Reading

Phil Harvey has been a Light Reading writer and editor for more than 18 years combined. He began his second tour as the site's chief editor in April 2020.

His interest in speed and scale means he often covers optical networking and the foundational technologies powering the modern Internet.

Harvey covered networking, Internet infrastructure and dot-com mania in the late 90s for Silicon Valley magazines like UPSIDE and Red Herring before joining Light Reading (for the first time) in late 2000.

After moving to the Republic of Texas, Harvey spent eight years as a contributing tech writer for D CEO magazine, producing columns about tech advances in everything from supercomputing to cellphone recycling.

Harvey is an avid photographer and camera collector – if you accept that compulsive shopping and "collecting" are the same.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like