← Back to Posts

IT meets telecom: Peter Rysavy on Kubernetes, O-RAN and the next five years

Sean Roberts•December 2, 2020•3 min read

The 5G conversation in Washington is about spectrum and the Pentagon. The one I wanted to have with Peter Rysavy of Rysavy Research was about software. A few startups have begun describing something they call enterprise 5G, built on the same open source stack that runs modern data centers. I asked him whether it has legs.

Absolutely. This is IT meets telecom.

For most of the industry's history, he explained, a mobile network was a huge system built by a small number of vendors out of proprietary boxes. The networks followed industry standards, 3G, 4G and now 5G, but by and large one vendor supplied all the building blocks. Virtualization began to change that, by implementing more of those functions in software. Containers take it a big step further.

I pushed the idea as far as I could. Kubernetes already orchestrates software across laptops, Amazon, Google and private data centers, and projects like Istio, the service mesh Google backs, manage traffic between services wherever they run. If that orchestration layer extends over 5G, to networks of sensors, what does it mean for the long-running argument about the government sharing spectrum with commercial carriers?

Peter, sensibly, set the spectrum question aside as a separate issue. Orchestration is not separate.

These networks are huge. They have many different functions and many different interfaces. So bringing new capabilities, new features online that involve a lot of different devices and programs within the network, the management is essential.

The wireless industry's answer to that so far is ONAP, the Open Network Automation Platform, an open source project for running large virtualized networks. Next to it is O-RAN, a set of open interfaces between the major parts of the radio access network, so the radio side can be virtualized and the pieces can come from different suppliers. Kubernetes, O-RAN and ONAP together, he said, are transforming the wireless network. It is a long-term effort, and the networks are going to look very different five years from now than they do today.

There is already evidence for that. Rakuten launched a mobile network in Japan this spring built largely in software, and Dish has said its new 5G network in the United States will use O-RAN. The C-band auction opens next week, and the carriers bidding billions for spectrum are also moving their networks toward the way cloud providers run theirs.

I have watched this happen once before. A decade ago the data center ran on proprietary appliances, each vendor's box doing one job its own way. Then the functions moved into software on commodity servers, open source took over the orchestration layer, and the companies that sold boxes lost control of the stack. Peter's description of telecom matches that history step for step. The difference is that a mobile network has a radio at the edge and a regulator watching it, which is why the open interfaces in O-RAN matter more than any single product.

So here is a prediction worth dating. Peter said five years. That puts the check at December 2025: by then, will a typical mobile network be run as software from several suppliers, or still as boxes from one?

From Lincoln Shorts, the interview series I hosted for Lincoln Network, recorded 2 December 2020.

Share this post