Operators Need to Automate Networks; AI Makes Automation Smarter

Carriers are pushing automation in their radio access networks but “need to be smart” about how they deploy it, Adam Loddeke, AT&T assistant vice president-RAN technology, said Monday during an all-day IEEE virtual workshop. Speakers also discussed how AI can supercharge automation.

Carriers must minimize any risk automation poses “given how much we rely on it as we scale it into the network,” Loddeke said. Carriers are looking to move to open RAN, he noted. Smaller vendors may have new ideas, he said. “It’s important for us to have a vibrant ecosystem that supports multiple vendors,” which allows “innovation and technology that otherwise might not be present,” he said.

Carriers are moving toward near real-time RAN intelligent controllers (RICs), which will mean the “near instantaneous ability to make changes to the network,” but there are “challenges,” Loddeke said. “There’s a mix of levels of support in the ecosystem as it is today,” he said. AT&T wants a single platform that can support multiple vendors, he added. “We want something that’s somewhat standards-based -- we don’t want something that’s unique to AT&T.”

Today, each time AT&T upgrades the element management system in its network, it requires “a significant amount of internal development work to support all of the internal systems,” Loddeke said: “When we pivot to a single common management system it becomes much easier to maintain the lifecycle and to support the automation necessary within that platform. … It makes it much easier to onboard future vendors.” There’s no question that disaggregation of the network adds complexity and will require more network automation, he said. “Without automation we’re going to run into trouble -- there’s no doubt.”

Everyone mentions AI and automation, sometimes as if the terms were interchangeable, said Rajarajan Sivaraj, director-ORAN architecture and standards at Mavenir. “Automation is, as the name suggests, automating a complete workflow,” while AI makes automation “more intelligent,” he said.

Today’s densified networks are complicated, Sivaraj said. Automation allows “consistent operations at scale,” with an ability to monitor the network and make changes when needed, he said. “AI adds a layer on top of this automation,” and “intelligence” into how problems are solved. AI “provides for deep insights about the network from large amounts of observability data” allowing operators to make “more informed decisions,” he said.

AI can correlate different types of data, at various levels in the network, and explain the “root cause” of “observed network behavior” to make automation work better, Sivaraj said. AI can also forecast network behavior and recommend configuration changes that "can be automated for percolating down to the network functions,” he said. Moreover, AI can provide a digital trend in the network before automation tools make changes to the network’s configuration, he said.

Innovation and automation have moved from “nice to have” to something needed for networks to survive and grow, said Ramesh Nagarajan, Google global head-network modernization, telecom business unit. Developers still tend to throw things “over the fence” and “then they move on to the next big feature or product,” he said. Operators “are constantly juggling how to manage” new features that are delivered to them, he said.

Carriers are responsible for managing critical networks their customers rely on, Nagarajan said. “Developers want to go as fast as they can,” he said: “Operators are looking to kind of slow things down as much as possible to reduce risks.” This is potentially “a pretty horrible” situation, which isn’t designed for harmony and doesn’t “scale well,” he said. Carriers have a “maniacal focus” on preventing failure, which “slows down the pace of innovation and delivery,” he said.

Nagarajan called for a change in operator mindset, with a willingness to accept failure “as normal” while designing systems that can accommodate a certain level of failure. Operators also need to break down “silos” between the developer and operations teams “so we have a strong collaboration culture.”

AI needs to be explainable and transparent about how it works, said Arvinder Anand, Ericsson head-architecture, technology and end-to-end 5G solutions. Enabling explainable features in AI “is crucial in creating trust,” he said. Understanding the actions of an AI system and the rationale behind them is essential, he said.