Seattle-based Opanga Networks says it has found a way to minimize wireless network congestion and improve traffic flow by using machine learning algorithms in the network core. The company’s software is being used today in 3G and 4G networks to make them more efficient without requiring more spectrum

Basically, Opanga’s software platform is deployed in the packet core. The software, which can run over existing hardware or new COTS hardware, uses machine learning to identify what Opanga calls “elephant flows.” These are large users of capacity. According to Opanga CEO Dave Gibbons, about 3 percent of data sessions on a wireless network generate about 60 percent of traffic. Those 3 percent of data sessions are “elephant flows.”

Gibbons said that Opanga’s software manages elephant flows before they hit the access network. “We don’t try to change the data flows but we manage the way they flow into the radio access network,” he said.

That’s where the machine learning technology comes in. By using machine learning, Opanga can monitor the flow of the traffic (which Gibbons says is primarily video) in the core so that it doesn’t saturate the base stations. Because many wireless operators are still using a standard packet core network, Gibbons said that Opanga’s technology can be deployed in a network with a standard core technology, but when it’s deployed in a virtualized evolved packet core (vEPC) the software can be deployed more rapidly. “The vEPC is a turbo charger,” Gibbons said.

“The software manages the traffic and leverages the move to the vEPC by deploying these algorithms,” he added. “That improves the performance of the RAN.”

And while some operators are looking at virtualizing the RAN through various initiatives including xRAN, ORAN, and vRAN, Gibbons said that Opanga's software platform can work with both the traditional RAN and the vRAN. "It doesn't matter if the RAN is hardware or vRAN," he said.

Gibbons said that the company’s software is applicable in 3G and 4G networks today and will be relevant to 5G networks of tomorrow. But he argued that many operators are overlooking what can be done today to improve network capacity without the need for 5G.

“There is so much talk about 5G and massive MIMO [multiple input multiple output] technologies. They are great but what about leveraging the vEPC to do intelligent things in the network core?” he asked. “That’s a huge opportunity.”

Gibbons said that the company’s technology is deployed in multiple networks around the globe including in the U.S. But he also admitted that as a small firm the company does rely on its partnerships with the larger OEMs to sell its software to wireless operators. “The traditional OEMs recognize that delivering software intelligence to densify your RAN will be important to the revenue lines,” he said.

The privately held company was started back in 2005 by some wireless industry veterans. However, the company stayed under-the-radar for several years focusing on research and development.

CEO Gibbons previously was the CTO of AT&T Wireless’ OFDM fixed wireless division, and Vesa Raudaskoski, Opanga’s VP of business development, worked for Nokia’s infrastructure division in the early 1990s. Gibbons and Raudaskoski also founded Eden Rock Communications, a self-organizing network (SON) technology firm that Nokia acquired in 2015. Opanga’s board of directors also includes high-profile wireless veterans including Cole Brodman, the former CTO and CMO of T-Mobile, and Bill Owens, the former CEO of Nortel Networks.