11:25 AM -- SANTA CLARA, Calif. -- Open Networking Summit -- OpenFlow is a Layer 2 technology, and software defined networking (SDN) is a discussion about Layers 2 and 3. It's switch-and-router stuff.
But some of the technology's biggest fans might be working at Layer 1. If SDN takes off, it might be great news for optical transport.
I hadn't thought of it that way until talking to Ping Pan, an architect at Infinera Corp., Tuesday afternoon. For him, the appeal of SDN is that it lessens the importance of Layer 3 intelligence.
That's because SDN turns the network "sideways," as he puts it. Rather than let Layer 3 decide where packets go, the network now gets told to get this much bandwidth to that application by any means possible. It doesn't matter what kind of box executes that command.
"Now all you need is northbound API [application programming interface]," Pan says. "Why would you put a router in front of us? That's not what Google's doing. They're saying they don't need a router."
Google, in fact, is running OpenFlow on 100 percent of its internal backbone, a project that was the subject of Tuesday morning's keynote. (See Google Uses OpenFlow Massively.)
I don't think Infinera is the only optical company that sees this. While Pan and I talked, someone with a Ciena Corp. badge walked past. Come to think of it, I also spent part of lunch with Fred Gruman, principal product planner at Fujitsu Network Communications Inc., who came here to learn more about what's been happening in SDN. It wouldn't be a stretch for him to have reached the same conclusions.
What Pan finds exciting is the prospect of some of the revenue coming back down to the transport layer. It would break the trend of the better margins always being found further up the stack, letting the transport side make better money on all this bandwidth growth.
He puts it this way: The router guys are scared of SDN, he thinks. Software people are happy about it, but in a general way, they don't yet know how SDN is going to be applied. But the transport guys should be ecstatic.
The downside is if more companies follow Google's "none of the above" route by using generic switches. (Google's are homegrown, but that's an anomaly.) Still, carriers aren't going to stop needing optical transport. Maybe SDN is a chance for them to really converge some network layers.
— Craig Matsumoto, Managing Editor, Light Reading
