One thing on the road map for folks at AT&T this year is finding a home for Open Architecture for a Disaggregated Network Operating System (dNOS), a vision for which it released last November in hopes of getting hardware and software vendors, open source developers, other service providers and more behind it.
AT&T put out a white paper last year providing an overview of dNOS, and the subject is already generating interest. “Early responses indicate that we shouldn’t have a problem building a community around it,” Chris Rice, senior vice president of AT&T Labs, Domain 2.0 Architecture and Design, told FierceWirelessTech.
What exactly is dNOS? Rice explained that the way to think about it is there will be places where AT&T wants to use white boxes, which could be switches or routers, but they’re boxes that it will use in its network to perform specific functions that need some kind of operating system associated with them.
“What we would like to be able to do is have that done in a very consistent way so that it easily interfaces and connects to our network,” he said, adding that his company has gone through and looked at what would be required to make it happen.
But it’s not just for AT&T’s sake; the operator believes it will apply to other service providers as well, including potentially its rivals in the U.S.
Rice acknowledges that efforts like this have been tried before; some remain in existence and others went by the wayside for lack of support. Examples of existing ones are efforts at Facebook, which created the Open Compute Project (OCP) and subsequent Telecom Infrastructure Project (TIP), Microsoft SONiC and Free Range Routing within the Linux Foundation.
“The goal is really to do something that will satisfy white box requirements for service provider needs,” Rice said, noting there was a feeling that previous efforts didn’t suffice. Those other efforts “were just not quite there for what we would need, and we’d like to put something out that is more fully encompassing of what our needs might be and we believe other service providers’ needs would be.”
OCP does a lot of referencing to hardware designs, but it’s not sufficient in the sense that software is needed to go with that, so “think of dNOS as the software that would then go with that hardware,” he said. Things in OCP can be used with dNOS, for example.
The idea is this will be open, shareable and best of breed. “We think that the area where we believe that folks can add the most value is in the application and protocol area,” he said.
And how does this relate to ONAP, the Open Network Automation Platform that combined AT&T’s ECOMP with Open O? Rice said to think of them as being complementary, but they’re two distinct and equally important initiatives. ONAP is the network operating system for the cloud; dNOS is basically the networking operating system for the white box.
Similar to what happened with ONAP, a prime home for the dNOS endeavor would likely be in the Linux Foundation, but there’s nothing set in stone there and no announcement is being made.
ONAP is less than a year old and now has about 50 members, including a dozen or so service providers.
There’s nothing that says a company would need to participate in both ONAP and dNOS; it could do one without the other, but Rice anticipates more momentum building for dNOS: “I can’t speak for any particular operator, but I would expect many of them have the same need and expect them to be interested in something like this.”