SDN Technology

What Martin Casado Learned From His SDN Years

SANTA CLARA, Calif. -- Back in 2011, when the Open Networking Summit (ONS) was a 400-person gathering held on the Stanford University campus, one of the featured speakers was Martin Casado, a founder of software-defined networking (SDN) startup Nicira.

Casado, who became a VMware Inc. (NYSE: VMW) executive after Nicira was acquired, kept coming back to ONS, which became a gathering place for Silicon Valley's SDN community. But this year, he was back in a role as a venture capitalist, a general partner with Andreessen Horowitz, and his Tuesday ONS keynote was a look back at his SDN experience.

Some of his most important lessons from the startup years related to the realities of business. For example: As difficult as hardware and software development might be, the bigger hurdle is customer behavior. "Instead of dithering around on the platform, you want to get into market as quickly as possible to get that feedback and start that education process," Casado said.

Even then, if you happen to be targeting a slow-moving industry -- telecom, for instance -- you'll need to be patient. "This all comes down to the speed at which an industry moves," not the speed at which you can develop the technology, he said.

Casado did learn a lot about the networking world, too. Casado and fellow students at Stanford, working on projects that were predecessors to Nicira, wanted to apply software principles to the data center network. Their work was motivated by the fact that new features weren't available on switches until the hardware upgraded. Their thinking was: Why not separate the software and let it upgrade independently?

That's how servers are built, after all. But networking is different. Networking is "fundamentally about distributed state management," he said.

Moreover, networking comes in multiple flavors that don't go well together.

"I'm more and more of the opinion that it doesn't make sense to have a general SDN platform, because the problems are so different," Casado said. The WAN, the data center, the campus LAN -- it doesn't make sense to seek out a platform that suits all three, he said.

Looking back as a founder and forward as a VC, Casado noted that success is "overdetermined" -- a math expression referring to a set of equations that have multiple solutions, even infinite solutions. That makes it hard to tell what truly worked.

"You can make a lot of bad decisions, but a few key decisions drive all the outcomes," he said. "But unlike venture [capital], you can't take a portfolio approach to this. Unlike venture, you make these bets in serial."

He doesn't seem to miss that life. If anything, Casado seems to relish his new opportunity to "go broad" and chase many problems at once.

"How many decades does one have in a life? I've already used one."

— Craig Matsumoto, Editor-in-Chief, Light Reading

Sterling Perrin 4/5/2017 | 10:26:57 AM
SDN Controllers Very interesting comments from Martin Casado. On SDN controllers, it looks like 3 buckets have emerged just as he has described: 1. DC controllers; 2. enterprise WAN 3. carrier WAN.

Some suppliers have consolidated those buckets across two product lines: 1. a DC/enterprise WAN produdct and 2. a carrier WAN product.

But the carrier WAN seems to be its own unique product in all cases. It also relates to his comment about how telecom moves slower. I like a quote from a tier 1 North American operator exec I interviewed on SDN last year who said: "Technology is easy. People are hard." Introducing technology at right speed for the organization is an important problem that needs to be solved for telecom operators.


Sign In