Exploring SD-WAN Technology

Fortify Your Data podcast. Published 20 May 2019. Recorded in Cleveland while I was running WAN Dynamics.

An early one. In 2019 SD-WAN was mainstream enough to have a Gartner Magic Quadrant and new enough that I was still being asked to explain it in every first meeting. This conversation is a good snapshot of what I was saying then, and most of it has held up.

How I got here

Dial-up support at Stratos Internet Group, the first $10-a-month dial-up in Cleveland, where some callers deliberately broke their settings so they would have someone to talk to. Then into the telco side, where I helped build a $6 million DSL network across Cleveland, Chicago and Columbus, put about ten customers on it, and watched the vendor take the gear back when the money ran out. Late-90s telecom taught me what not to do. Thirteen years at a regional service provider took me from junior engineer to product development and solution selling, and when that company was acquired, we launched WAN Dynamics with the products we had been building.

Discovered by selling against it

I found SD-WAN in a competitive deal. I was pitching MPLS and IP VPN, and the other guy in the room was selling something so new it was not even called SD-WAN yet. My sales rep walked out saying "we've got this." I walked out thinking it was way cooler than what we had. A network that measured every path continuously and steered applications across them dynamically was something I had not seen in twenty years of networking. The acquiring company did not get it, so we built the company around it instead.

The short definition, and the misconceptions

An application-aware overlay network built on top of whatever transport you want: dedicated internet, MPLS, 4G, fixed wireless. The links become immaterial. The biggest misconception was already the one I would spend years correcting: that it will save you money. CIOs called wanting a quote that would cut the WAN budget 20%. Usually the spend is neutral and what you get is a better-performing network and a more productive workforce. Early adopters were large retailers who needed the transaction to always go through, then professional services firms moving big files who liked bonding links together. Banks were last, holding MPLS as a security blanket. Healthcare used it for the satellite offices that had always been hard to tie in.

What people learn about their own networks

The story everyone remembers from this episode: a manufacturer brought us in for a demo, we dropped a box on the network, and the dashboard showed 60% of their WAN traffic was YouTube. Hundreds of workstations, everyone building a playlist for the day, and constant performance complaints nobody could explain. One policy capped it at 10% of the link and pushed it onto the cheap cable circuit, and they asked where to sign before they looked at the price. BitTorrent rate-limited to 0.01% and file-sharing apps that policy had supposedly banned were the other usual surprises.

Where it was going

I predicted SD-WAN would stop being a standalone product and become a feature of edge computing, pointing at Dell's announcement of an edge box with VeloCloud built in and workloads on VMware, with the ability to move applications between edge, private and public cloud. I said "SD-WAN" would eventually just be "WAN", the way cybersecurity became security. On the telcos: innovation used to come from Bell Labs, now it comes from Google and Facebook, and SD-WAN pushes carriers further toward being the dumb pipes an overlay rides on. On security: encryption everywhere is a genuine gain, especially the reminder that your MPLS provider is not encrypting your traffic for you, and the caveat was that the proprietary tunnelling protocols had not yet been battle-tested. And on AI, which the host insisted on: business intent driving policy existed in a small way, and the interesting future was anomaly detection and dynamic optimization feeding back into the network. That part arrived on schedule.

Watch on YouTube →