Building Next-Gen Network Is Difficult Even Without Legacy Issues: Dish Wireless

Dish Wireless wanted to do something unprecedented: design, build and deploy the world’s “first of its kind 5G network” in only three years using the public cloud, Eben Albertyn, Dish Wireless executive vice president-chief technology officer, said during an RCR virtual conference Thursday. Several experts mentioned the growing security and other challenges facing carriers in a virtualized-network world.

Dish's goal was competing as a fourth national carrier in the U.S. market “using best-of-breed technologies,” Albertyn said. Reaching initial FCC network benchmarks required the right people and a “culture of innovation,” he said. Even without the complications of building on top of a legacy network, a virtualized network has been challenging, he said.

“The amount of technological breakthroughs and achievements we’ve plowed through … is quite staggering,” Albertyn said: “It was and continues to be challenging to make sure that the carrier-grade network can function in the way that we’ve been able to function.” Dish has found it can’t view any piece of technology as something it won’t try or that it will discard if it’s not working, he said: “It’s about the end objective. … We don’t have any holy cows in terms of technology.”

Not only must carriers worry about security, they have to protect new 5G networks that are “more open, very software-centric, very cloudified,” said Michela Menting, senior research director at ABI Research. “There’s a big learning curve for everyone involved.” To date, more than 220 5G networks have launched and more than 30 have a stand-alone core, she said. “We’re in the midst of a rapid transition.”

Open radio access networks have also made progress during the past year, Menting said. Carriers will likely move from cloud-RAN to ORAN “eventually,” she predicted. Moving to the cloud and ORAN means additional risks, Menting noted.

Carriers will put cloud hardware at the cellsites “and it’s going to be operationally far more complex” than today’s networks, said Amy Zwarico, AT&T director-cybersecurity. A single deployment will include hardware and software potentially from more than a single vendor, she said. “That’s the promise of ORAN -- I don’t have to use a single vendor in a deployment.” But that adds complexity, she said.

One big change is that vendors must "address the complexity of aligning their operational and their security practices to be able to provide very frequent security updates and patches,” Zwarico said. Providers aren’t used to rapid changes in their networks, she said. The threat service will be wider and more complex, she added. Moreover, the continuous integration and continuous deployment (CI/CD) pipelines will be more complicated. It’s no longer a matter of installing and plugging in a single baseband unit. “You’re going to have a full CI/CD pipeline to deploy,” she said.

Other speakers said AI can help providers address complexity.

The term AI is “quite abused,” said Prakash Sangam, principal at Tantra Analyst. “Everything and everyone seems to be talking about it,” he said. “Undoubtedly” AI “has a huge role to play in the complex task of managing large, dynamic telecom networks,” he said.

Complexity is the biggest challenge providers face and will become more of an issue as networks are disaggregated, said Ajay Rajkumar, principal member of the technical staff at AT&T. The automation that AI will lead to is “the biggest opportunity” available to providers, he said. Networks started using machine learning years ago and its use is growing with 5G, he said.

Making networks more automated requires reliable AI, Rajkumar said. “The cost of small mistakes or one single mistake can be huge -- these are critical networks.”

Carriers have a growing amount of data about how their networks are functioning, said Chris Murphy, regional chief technology officer at Viavi Solutions, a test measurement company. The data needs to be collected, harmonized and cleaned, he said. “It’s not always easy to bring the data together to come up with a coherent view of what’s going on, so we can perform advanced analytics,” he said. End-to-end automation is “what we’re ultimately aiming for,” he noted.