Software-defined networking (SDN) has certainly come a long way, but a closing panel at the Open Networking Summit (ONS) offered a dose of realism, noting that many users still feel out to sea with the technology.
“What SDN needs now, most, is customer success stories. Large customer deployments — and not Google. Google’s fantastic, but they’re not the mainstream,” said Prodip Sen, Verizon’s director of network architecture, during Wednesday evening’s closing session.
Sen and other panelists were optimistic about SDN’s prospects during the next 12 months. None of them were arguing that SDN is experiencing setbacks.
But looking ahead to the 2015 ONS, panelists agreed that the audience is going to want evidence of how SDN has helped customers in the real world.
That’s particularly true for the enterprise. Google-like examples of SDN in large data centers abound, and service-provider experiences and plans made up a good chunk of the ONS sessions. “But regular enterprise data centers, I think, still are struggling with what value SDN brings to them,” said Vijoy Pandey, CTO of network operating systems and SDN at IBM.
His suggestion for 2015? “Getting somebody on this stage from an enterprise data-center point of view.”
SDN is a new, complex concept that’s not easy for those who haven’t followed it for the past four years. Pandey noted that vendors “talk about 10 diverse things” and lump them together as “SDN.” That doesn’t clear anything up for the enterprise customer.
“A lot of the vendors need to step up and make it more consumable, to show it’s easily deployable and has some value,” said Guido Appenzeller, a Big Switch Networks executive and former CEO.
Pandey pointed to the nine proofs-of-concept for network functions virtualization (NFV) that the ETSI Industry Specifications Group (ISG) has published. Those examples articulate NFV’s benefits — and they’re demonstrable; ETSI PoC exhibits were all over Mobile World Congress last week.
“For the SDN side of things, for the enterprise data center, something similar needs to be done,” Pandey said.
SDN will also need to offer the enterprise more support and a simplified marketing message, because as the technology spreads, it’s going to reach customers that don’t want to, say, take an open-source controller and mold it to their liking. That shift toward the mainstream is one reason why Big Switch changed trains to pursue a bare-metal strategy.
“My favorite customer at the moment is an apparel retailer in California. They’re a very innovative company, but they’re not the same as a Google or an Amazon. They don’t have the development team,” Appenzeller said. “Having a programmable platform is not so important. They’d rather have a turnkey platform.”
(Big Switch’s other motivation, as Appenzeller noted, was the lack of full-fledged OpenFlow support from commercial switch vendors. But that’s another topic…)
(Some) Carriers Love SDN
Carriers, meanwhile, have really started running with SDN, due to their eagerness to match the agility of Google, Amazon, Facebook, and other hyperscale players. We heard this in John Donovan’s ONS talk about his ambitions for AT&T.
“The urgency is there to do something. All of us face competition at every level in every domain we operate. We cannot afford to do nothing,” Sen said.
But being new, SDN still has its gaps.
“A lot of development has to happen in that space,” Sen said. “We need tools there. We need operations support things there that allow us to manipulate our network as a whole. … To do that, we need all sorts of orchestration, automation — things we don’t have right now.”
(Photo, left to right: Pandey, Sen, Appenzeller.)