What if nearly every network operation could be predicted using machine learning and AI?
Ericsson launched its Operations Engine in January to assist communications service providers with moving away from incident driven and reactive operations to predictive and preventative operations. The engine, which leverages machine learning and Artificial Intelligence (AI), is an end-to-end managed services operating model that reimagines network and IT operations, network design and optimization and applications development and maintenance.
At Mobile World Congress Los Angeles, Ericsson’s Head of Network Managed Services Bradley Mead walked RCR Wireless News through the formation of the Ericsson Operations Engine, explaining that the company basically started from scratch, or in Mead’s words, “with a clean sheet of paper.”
Put simply, Ericsson began thinking about what incident management would look like if nearly everything could be predicted using AI and machine learning.
“We redesigned all the standard processes that you normally think about when it comes to incident management processes,” Mead continued. “And then we said, ‘what are the new processes you would need?’”
Ericsson worked on answering that question for over a year.
Amazingly, this work began before Ericsson confirmed that the processes being redesigned could even be predicted at all. “It turned out,” Mead reflected, “we were pleasantly surprised that even some things that we didn’t think could be predicted, can be. And quite accurately. That added fuel to the process redesign.”
According to Mead, another game changer offered by the ability to predict an incident, rather than simply react to it, is the ability to measure customer experience properly from end-to-end. “We’ve talked about customer experience for many years, but as an industry we’ve been working without the right capabilities to easily measure it,” he said.
But now, with solutions like the Ericsson Operations Engine, that ability exists. “And that’s exciting,” Mead added.
According to Mead, Ericsson has always been a believer that even when the technology within a network is siloed, the best way to engineer and operate a network has always been end-to-end. “But now, you have no choice,” Mead explained. “You actually need to look at the network holistically and you need to engineer and operate it in a very end-to-end, holistic approach. That’s something that is really different.”
A flatter, un-siloed network architecture like what Mead described allows operators to go to market with new services and to differentiate the services based on customer service levels. “They can provide more unique and flexible services and adapt and change over time depending on customer demand and feedback,” he said.
Another pleasant surprise for Ericsson was that many of the new processes they developed could actually be applied to legacy systems. “Once we came up with the new processes and defined the new skills necessary to run them, we didn’t think you could apply much of it to what exists today because everyone has legacy networks that need to be migrated over time,” he explained. “We thought we might have to keep some separation and do things the old way for a while before migrating the systems over. What we actually found is probably about 60 or 70% of the new processes could be applied to the legacy processes and technology.”
The biggest question one might have, Mead said, is why do this? Why use machine learning and AI to predict and prevent network issues?
“At the end of the day, it’s going from reactive to predictive. That is what everybody wants,” he said. “If you look at the workload demand on a network operation center it comes from two sources: customer complaints and alarms and faults in the network. All reactive in nature, so using AI and machine learning to predict these issues is amazing.”
“Even on the reactive side,” he continued, “as you look to optimize networks, the complexities in terms of number of nodes and sites of a network and the layers of technology, there is too much data, and you can’t throw enough people at it.”
Mead said Ericsson is currently working with a number of key customers to implement their Operations Engine and is in the middle a number of big transformations. One project is with a large operator in Europe that has five operations in different countries. Other projects include working with an operator in the U.K., one in India and one in the U.S.
A demonstration of the Ericsson Operations Engine on the show floor revealed just how much predictive information an operator is able to receive about its network. A live map displayed all of the network nodes and indicated which nodes were likely to suffer an incident and how much time the operator had to respond to the potential problem before it became an active alarm.
In addition, the solution provides tremendous flexibility by allowing an operator to select desired business outcomes, such as grow revenue, improve customer experience and optimize cost. Once the system understands the operator’s business needs, a number of capabilities are displayed that address the service and performance metrics related to the outcomes selected.
As the conversation came to a close, Mead shared his personal opinion about the direction the telecom industry is going: “This is more [my]view, I guess, and not necessarily Ericsson’s, but I think the adoption of 5G is going to go faster than any other technology that we’ve seen in history in the telecoms industry. I don’t think we’ve had a change like this since we’ve changed from GSM to wideband CDMA back in 2002-ish. It’s that magnitude level of change, and then some.”
He continued, “What makes it different this time is that all the planets are aligned. The devices are there and are more in sync with the technology than they’ve ever been. The demand is there. And the operators need it.”
The post Machine learning, AI aids network incident prediction, prevention appeared first on RCR Wireless News.