OPNFV update: OPNFV is just a year old, but more than 700 people attended its inaugural OPNFV Summit in Burlingame, Calif., this week. The open source group has assigned itself the task of creating a full stack for network functions virtualization (NFV) that can be deployed by telcos.
While the goal is daunting, the overall mood of participants was eager and optimistic. However, there’s a lot to sort out, and not just the technical work. Thursday morning’s panel discussions had little to do with code and lots to do with defining the direction of OPNFV.
“Three years ago there was no NFV. Back in 2013 users were asking: What is NFV?” said Prodip Sen, the CTO of NFV at HP and chair of the OPNFV board of directors. “The conversations were largely around cost reductions being the business driver. Now the conversation is about agility and elasticity.”
As to why OPNFV was even formed when there were already so many open source groups, Sen said NFV “brings stuff from all the other open source efforts. When we started, we thought we were building one platform, but it’s really a framework where you have a lot of choices.”
Chris Wright, chief technologist with Red Hat, said network users’ expectations are being set by Internet companies where orders are self-service and immediate. “Tomorrow’s network is software-centric, with apps that scale out on-demand,” he said.
However, network environments are much more complex than many of these Internet companies that have raised the bar. “We’re looking for a set of consistencies or templates or abstractions so we can use tooling to replace the artisan, handcrafted snowflakes that make up a hardware environment,” Wright said.
“We’re trying to bring those pieces together,” he said.
Sen made it clear that OPNFV’s work “is not a product release.” He said it’s about the structure. “The release cycle is based on stability of the system, testability of the system, how well it’s integrated. We shouldn’t be spending a whole lot of time on designing features.”
Neela Jacques, executive director of the OpenDaylight Project, said at a high level there are several ways to approach software defined networking (SDN) and NFV. He said carriers have at least a three-year depreciation cycle, probably more like 10 years, for their existing network hardware. So they may decide to transition by making their existing networks better with some virtualization.
“The other side of that is where we do have the ability to build a whole new network,” said Jacques. You can do NFV without SDN and vice versa, he noted, but if you put them together you have “the perfect story.”
“Some of it is timing alignment,” said Jonathan Bryce, executive director with the OpenStack Foundation. But he recommends that telcos and even enterprises should be planning ahead because “the only way you can respond quickly is to have software.”
Then There’s the People Element
OPNFV will likely experience the same problem with corporate cultures. Bryce recommends that companies use a carrot-and-stick approach. The carrot is education and also cheering on the new technology as a way to get changes done more rapidly. The stick can be in the form of policies such as not approving new hardware without sign-on by a VP, for example.
Besides the end users, OPNFV has its own people concerns. It needs more developers. Currently, most of its developers come from its member companies. But it wants many more talented coders.
At a media lunch at the conference, Jacob Loveless, CEO of Lucera, suggested that OPNFV find a way to entice programmers from startups. He said they would come if participating provided them a way to hob-nob with potential investors.
OPNFV isn’t waiting very long for its next conference. It’s planning its next OPNFV event in Europe in about six months.