NFV (Network functions virtualization)

AT&T Working on Home-Grown SDN Controller for Later in 2014

AT&T is plunging deeper into the world of software-defined networking with plans to develop its own controller for network bandwidth allocation and other services this year.

At an "Innovation Showcase" in New York City Friday, Marian Croak, AT&T Inc. (NYSE: T) SVP of Domain 2.0 architecture and advanced services development, revealed that the operator is working on a "controller" to manage network functions and services as Ma Bell moves to run its network on "low-cost hardware" with the functionality abstracted into software. (See AT&T Spotlights Early SDN Efforts.)

"We're building this programmable controller now. I think you'll see it later this year," Croak said. It will enable AT&T to better manage and reroute network capacity, amongst other services.

This wouldn't be the first time AT&T pioneered its own SDN developments. The company has been running an early version of SDN, known as intelligent routing service control point, or IRSCP, in its core network for six or seven years now. That technology, which separates the data and control planes and allows for distributed control of routing, is now powering AT&T NetBond, a service that lets enterprises connect to the AT&T cloud via secure links that appear as nodes on their MPLS-based virtual private networks. (See AT&T Spotlights Early SDN Efforts.)

The eventual aim is to let AT&T's customers tailor their networks to their own needs. "As customers of AT&T, you'll be able to design your own network," Croak said. This "user-defined network cloud" needs maximum flexibility in use of network resources.

There is a lot of work to be done, however, between now and AT&T reaching that lofty goal. To that end, AT&T has been refocusing a lot of its research teams on NFV and SDN work, Greg Bond from AT&T Labs told Light Reading at the event.

"The controller doesn't exist yet," Bond told us. He says that AT&T is working on multiple software-defined projects in the labs, as well as with established vendors and startups.

He stresses that AT&T isn't wedded to any particular specification for projects but has employed some for ease of use and so that partners have a frame of reference to work with. For instance, AT&T has used OpenFlow for data center bandwidth balancing.

Bond says that the new controller will pull from work being done across AT&T Labs. "We're working on three levels of orchestration," he says. The operator is using the European Telecommunications Standards Institute (ETSI) standard as an SDN/NFV template for now.

Bond admits that AT&T is working on a tight schedule to pull together its ambitious software-defined plans. "We're working really hard on this," he says. (See AT&T Reveals Audacious SDN Plans.)

Still, Bond is cheered by the way AT&T has steered its ship resolutely into software waters. "I'm a software guy. So it makes me happy."

AT&T believes its early work in SDN, through the IRSCP and other efforts, gives it an edge in deploying virtualization, according to Chris Rice, vice president, advanced technologies, at AT&T Labs.

"We have seen how SDN can be deployed at scale in a wide-area commercial network," Rice told Light Reading. "We are some of the early pioneers of this stuff, and there are other things we've done that people haven't seen yet."

— Dan Jones, Mobile Editor, Light Reading

DanJones 5/20/2014 | 6:22:27 AM
Re: The proof is in the paper At the event, Chris Rice briefly mentioned that AT&T is becoming a bigger user of -- and contributor to -- open-source software. He didn't elucidate further, so take as you will.
DanJones 5/20/2014 | 6:20:18 AM
Re: The proof is in the paper My guess is that they're not seeing a system they can buy that will do all they want it to do. That seems to generally be the reason that AT&T will generally get involved in dense software development.
t.bogataj 5/20/2014 | 2:16:50 AM
Re: The proof is in the paper An early version of SDN? The same you can say for T-MPLS, MPLS-TP, PBT, or PBB-TE: they all excluded the complex control plane from network elements. A nice article, but nothing revolutionary.

sam masud 5/19/2014 | 4:49:48 PM
Re: The proof is in the paper Other than saying we can build one too, is this ATT's way of indicating that it's not overly impressed with the work of the OpenDaylight Project or any other vendor developed controller (e.g. Cisco, Juniper etc.). Don't understand why ATT going down this road...
desiEngineer 5/19/2014 | 3:46:34 PM
Re: The proof is in the paper Ha!  What's their motto: Rough consensus and running paper?  I know I'm getting my standards bodies mixed up, but surely ATT isn't in the business of developing SDN controllers, and their objective must be to seed the community with a concrete way forward, not a 7-year-old paper that was written before the days of SDN (and VNFs which, though not necessarily related, are a great enabling environment for SDN).

Carol Wilson 5/19/2014 | 1:57:01 PM
Re: The proof is in the paper I asked that question and was told it kinda already is in that the paper is public and they won an award from DARPA for this work. I don't know that they are making a specific contribution to the ONF, however. 
desiEngineer 5/19/2014 | 1:20:31 PM
Re: The proof is in the paper Amazing stuff!!  I've been hearing about IRSCP for some time now.

Routing, QoS, redundancy, policy, DDoS, VPNs (aka overlay/underlay management), looks like BGP can manage it all.  SDN management was staring at us and we didn't realize it.

Will IRSCP be an open source contribution from ATT Labs?

Carol Wilson 5/16/2014 | 5:19:02 PM
The proof is in the paper Here's a link to the paper written by AT&T and UC-San Diego in 2007 on what is essentially an early version of SDN: 

Sign In