Network functions virtualization (NFV) is marching straight into open source territory, as a lot of the key players behind the concept are supporting the newly launched Open Platform for NFV Project (OPNFV).
Run by the Linux Foundation, the same entity organizing the OpenDaylight Project, OPNFV is launching with the support of 17 platinum members — including major carriers and vendors, such as AT&T, China Mobile, Cisco, and Ericsson — and 19 silver members. It’s a strong showing of support for the idea of making NFV as open-source as possible.
OPNFV is kicking off this week with a hackathon being held in Silicon Valley on Oct. 1. The group is aiming for its first code release to be in the first half of 2015.
NFV itself is looking inevitable. Driven by the large carrier customers that are demanding it, NFV has developed rapidly since being dreamed up in 2012, and there’s talk that the ETSI Industry Specifications Group (ISG) might respawn after its charter expires in January.
But outside the core ISG work, NFV is peppered with separate efforts from different vendors and carriers, sometimes creating partnerships that aren’t connected to the whole. “There are a lot of pairwise NDAs going on, so no one can share. You have a lot of the same conversations over and over,” says Margaret Chiosi, a distinguished network architect at AT&T Labs who’s been elected as the OPNFV’s first president.
OPNFV, then, is about speeding up the development of NFV by putting all of the common pieces into an industrywide, open source framework. It’s a lot like the motivation behind the OpenDaylight Project.
OPNFV’s goal is to create a carrier-grade reference architecture for NFV, formed of multiple open-source components. OPNFV will also work to tie this architecture to upstream projects that are already underway. It won’t develop any standards; that’s the purview of the ETSI group.
The initial focus for OPNFV will be around the virtualization layer, including NFV infrastructure and the management plane for the virtualized infrastructure. The northbound pieces that OPNFV would interact with include virtual network functions (VNFs), VNF managers, orchestration, and carriers’ OSS/BSS systems. (See diagram.)
Taking NFV Open Source
The OPNFV diagram should look familiar; it’s the “shape” of NFV as already defined by ETSI. But that concept, by itself, isn’t enough to fuel real-world applications.
“The goal, as I’ve told people, is that in the end, AT&T needs to focus on new services. And this platform is not something you can take out of the box or download it, and use it for new services,” Chiosi says. She’d prefer to have something more like a platform-as-a-service (PaaS), something ready-made that AT&T could build upon.
“Standards are great — if you already know what you want,” she says. “This is the Build side of NFV. We want to get to implementation faster as we experiment and innovate, and do it by coding, not by discussions, because discussions can take a while.”
“That’s what makes this a classic open source problem to solve. A lot of people are doing this in individual silos,” says Jim Zemlin, executive director of the Linux Foundation.
The restlessness that led to OPNFV began late last year, as the ETSI NFV effort was approaching its one-year anniversary. Other ISG members had the same concerns Chiosi did, and the idea of going open source started floating around. “I was skeptical, so I called up Jim Zemlin. He went through all these examples of how this was not as impossible as I thought,” Chiosi says.
Plenty of organizations were willing to host development of an NFV platform, but the Linux Foundation had the right experience in making an open source project work, Choisi says. That included all the financial issues that you wouldn’t necessarily think of: what costs might spring up, what kind of budget the project needs, and what fees and membership levels to establish. “We had no idea,” Chiosi says.
The legal side of open source, too, is new ground for many service providers. “We went to a lot of the operator community to explain these things, many of them seeing it for the first time,” Zemlin says.
The next six months will be spent on integration and testing of the open-source components that will create the baseline NFV platform, Zemlin says.
Before that, though, the OPNFV will hold its first meetings this week to settle on some base concepts, such as where the code should come from, which pieces should OPNFV borrow, and which should it develop itself.