How Does Open RAN Intersect with Edge Computing?

Cellular carriers, cable companies and hyperscalers are all investing in edge infrastructure, with market drivers including use cases ranging from residential cellular networks to industrial edge robotics and disaggregated open networks, said Bejoy Pankajakshan, EVP and chief technology and strategy officer for Mavenir.

During the Mobile Edge Forum virtual event, Pankajakshan said that a big shift in processing is coming. “If we look at how the applications are processed today, less than 10% of traffic is actually processed outside of a traditional data center coming from the operator,” he said. “If we look 10 years ahead, the predictions I’ve seen say that up to 75% of the processing would actually happen outside of traditional data centers.” Expected industries with expected growth in edge adoption include remote diagnostics and telemedicine, and industrial edge computing.

Pankajakshan sees Open RAN development overlapping with edge computing and an additional driver for it. In particular, he highlighted the role of the RAN Intelligent Controller, or RIC, which he likened to an app store for the RAN, where innovation can be implemented more agilely and faster than traditional single-vendor implementations.

RIC, he noted, comes in two forms: near-real-time RIC and non-real-time RIC. Both host applications: in the near-real-time RIC, these are xApps, while the non-real-time RIC has rApps. The difference between them is close to the control loops, Pankajakshan explains. “If you were implementing an application that required the control loop to run in less than a second, you would be running it in near real time on an RIC. … And if you have longer control cycles, then you would run it in the rApps,” he says.

This application-hosting capability makes the RIC itself a “flavor” of edge computing, he says, bringing intelligence deeper into the network — and potentially hosting additional third-party applications that leverage the available Open RAN network APIs.

“The reason this is so important for edge computing is that by moving some of the radio resource management and control functionality to the centralized edge infrastructure or smart radio controllers, you could now leverage third-party applications that leverage the APIs coming out of these boxes to build applications that could be used for enterprise or consumer use cases,” said Pankajakshan. “You could now host applications that are much closer together and have a per-user or per-device control mechanism.

He offered use cases such as traffic control or quality-of-experience optimization where edge computing could be applied to steer a given user towards better throughput or latency. “If a user is in a crowded area, you could optimize the user’s experience, even though the rest of the users in that location might not be having the best experience. … They improve the end experience that the business or the consumer could use.”

That means more opportunities for monetization, as well as a variety of human or machine user experiences that could be enabled by edge intelligence — and an opportunity for operators in 5G that was missed in 4G, where third-parties ended up developing APIs that were more consumable and simpler than operator open APIs and promoted adoption by developers. What does it take to actually implement such use cases? A lot of know-how in the RAN combined with machine learning, says Pankajakshan. “You would do detailed network modeling, you would build algorithms that can perform better, and then modify and retrain those models to get better optimized over time,” he explains. “They are building a library of smart network applications that can be consumed by the edge applications, and that could be a third party developer or the operator itself using that edge application to run their network better and provide a better experience for the edge -Applications to offer.

“As you go up the value chain here in terms of intelligence, you use better algorithms, you use better machine learning and AI algorithms here to give a model that can give better results as opposed to a closed system” , concluded Pankajakshan. “This is how we see Open RAN as well [the RIC] and the intelligence it brings to the network near the edge as one of the key building blocks for the operator to build intelligence at the network edge.”