"Interoperability will actually happen." For enterprises looking to leverage SDN more heavily, that 2014 prediction, given by the Open Networking Foundation's executive director, Dan Pitt, is good news. To find out why this could be the year interoperability becomes a reality, we dug into the obstacles vendors have been facing and where customers are most eager to see progress made.
Where the interoperability conversation started
Many of the challenges surrounding interoperability have been a function of time. Vendors simply needed to gain experience with OpenFlow 1.0 as a platform before tackling the task of getting everything to work together. "I think one of the impediments was we just didn't have enough products and enough experience out there," Pitt said. The conformance testing program was rolled out only last year, with developers originally targeting OpenFlow 1.0.1. And even though version 1.4 is available, ONF has advised developers that 1.3 is a stable target for testing this year.
Paul To, director of SDN and cloud at Spirent Communications and member of the Testing & Interoperability Workgroup at ONF, said interoperability has been a focus since the beginning. "There has always been a lot of interoperability testing during the early process for 1.0," he explained. The development of 1.3 was what really changed things up. "The 1.3 specification was stabilized and it was known as the specification that actually has a lot of the features the real world would use," To said. During the most recent PlugFest—an event that draws dozens of vendors together to test their OpenFlow-capable wares—in November 2013, things began to come together in a tangible way. "It was the first one where we saw a lot of vendors coming in with 1.3 support and passing a lot of the interoperability use cases," To said.
Market demand has also influenced manufacturers' efforts to achieve interoperability. Without enough SDN penetration within their networks, customers simply weren't clamoring for assets that communicated widely. That's changing now, according to Pitt. "What we're seeing now is market pull that demands these sorts of products and demands that they interoperate," he said. And AT&T's Domain 2.0 announcement, which Pitt described as possibly "the singular most important event in SDN in 2013," is also driving demand. He believes it signifies customer demand for products that separate hardware from software and forwarding from control. "When you have a sort of plug-and-play, modular world like this, you have to have interoperability to give the customer the freedom to choose best of breed for their need," he said.
To date, SDN interoperability development has remained focused largely within vendors' own product lines. This step forward shouldn't be seen as trivial. In fact, it signifies an evolution within the marketplace. "It's happening because just about every switch vendor has decided they need to also supply a controller," Pitt explained. Rather than develop silos around switches and software and controllers, vendors began to dive into creating a more widely interoperable set of solutions. "The switch vendors couldn't really sell their switches to customers saying, 'Just go find some startup OpenFlow controller to govern our switches,'" Pitt said. Instead, most started supplying controller code themselves, paving the way toward interoperability that could stretch beyond their own brand.
Where is interoperability headed?
One area in which enterprises will likely look for increased interoperability is "switching at the edge and switching closer to the core, where they're probably going to want to look at different depths for flow identification between the two, but they want to make sure they can send the right commands and identify flows and have them be findable," Pitt said. That focus may prove challenging, because complicated flow identification requires diving deep into the packet, something developers have found difficult. "We haven't had consistent ways of implementing that in silicon," Pitt said, adding it is a main focus of ONF's Forwarding Abstractions Working Group and Chipmakers' Advisory Board.
As SDN continues to move through the maturity spectrum, To said interactions with end-user organizations are on the rise, giving the industry a better understanding of customer needs. November's PlugFest had a strong technology focus, but, according to To, the next event "will take on more of a use case or use scenario focus." He anticipates the efforts will help administrators draw a clearer connection between development efforts up to now and how those will actually translate into the real world going forward. In addition, demonstrating interoperability in a lab is one thing, but proving how features work in the day-to-day operations of an enterprise could be something else entirely. "I think the industry will benefit from having more use-case based test scenarios," To said.
Pitt said he anticipates a "smooth and gradual evolution" for enterprise networks into increased OpenFlow capability. "A lot of vendors have provided OpenFlow support through a free software download into existing products," he explained. Among the benefits of future increased interoperability will likely be analysis and security, where Pitt said SDN can give administrators an advantage. "I think there is great hope among network operators of all sorts that [SDN] will be the best tool they've had for securing networks in a decade or possibly longer, because it is the most flexible and dynamic in responding to new threats."
Photo courtesy of Shutterstock.