Is There a Better Way to Do Software Defined Networking?

March 6, 2026·57 min·5 clips

Each router runs a full SDN controller and floods its local state to the rest of the network.

Meter sponsors the show. Ethan Banks opens from San Francisco, where he is attending the first North American Network Operators Group Gathering of 2026, and Scott Rabon sits in as guest co-host while Drew is away. The topic is software-defined networking. Alex Krenzel, a research engineer in the Systems Research Group at Google and a fourth year PhD student at UC Berkeley in the Netsys Lab, lays out a different way to run SDN than the usual centralized controller model. The usual model is described plainly. A controller or hierarchy of controllers computes network-engineered paths and pushes forwarding behavior into the network, which works, but also brings a lot of complexity. That complexity is the sticking point. Alex argues that the complexity of centralized SDN has led to serious network issues, which is why he wants people to look at distributed SDN instead. The conversation then gets into the mechanics. He describes a richer form of flooding that carries more information across the network, including demand data, so every router can build a global view in a link-state style. Scott presses on definitions. The group works through the difference between utilization and demand, with utilization tied to traffic on a link and demand tied to how traffic is entering the network at a router and heading toward destinations. Aggregation matters. Alex notes that the amount of state depends on how finely it is grouped, and that a router-level aggregate can keep the problem more manageable. Resources still matter. More compute and more storage make it easier to hold and process the changing flow information that distributed SDN would rely on. The state is not static. They keep circling back to the fact that flow information changes from second to second, which is part of why centralized controllers are hard to scale cleanly. Ethan does not pretend the idea is settled. He ends by saying distributed SDN is something he wants to do more homework on, and the episode closes by thanking listeners and everyone who spends time writing code, configuring gear, troubleshooting, optimizing, and caring about networks.

As heard by us

A practical look at centralized SDN, with a clear focus on operator tradeoffs and production reality.

Heavy Networking spends this hour on centralized SDN and the stubborn question of why an architecture built to simplify routing can still feel so involved.

Read the full review in PlayNext →

Why you'd press play

Hear a network-operator conversation about distributed SDN and why centralized controllers get complicated.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Heavy Networking