DENVER -- Light Reading Carrier SDN Networks -- Creating a standard version of SDN to which the entire industry can build is the telecom industry's next great problem, Verizon's Chris Emmons told a packed house here today.
"We've done this before, we did it with 4G, with 3G and with GPON," said Emmons, the director of network planning for SDN implementation for Verizon Communications Inc. (NYSE: VZ). "We know how to do this. We need to come up with a standard version and I don't care whether it comes from open source, from a true standards group or a forum. I don't think that matters as much as getting it right."
Having a standard version of SDN is "a prerequisite" to deployment, because Verizon needs to be sure that whatever it deploys can interoperate not only with other pieces of its network but also future networks, he noted.
Emmons, who followed Heavy Reading's Sterling Perrin to the stage, agreed with the analyst that there are almost too many different deployment options and evolving standards today for Carrier SDN. So while Verizon wants to move forward as quickly as possible with the strategy announced in April, there are still key issues to resolve. (See Verizon Builds Key Vendors Into SDN Strategy.)
And the Verizon executive agreed with Perrin on another point: That the product groups within network operators have different priorities for SDN than the network operations groups. "It's almost a dichotomy between what the customer-facing space wants and what the network group wants and we have to keep an eye on how those converge if we are going to get true end-to-end programmability in the network," he said.
Verizon and other network operators have tremendous incentive to tackle this issue because they are seeing demand for increased bandwidth without corresponding increases in revenues. "Consumers of that bandwidth aren't necessarily willing to pay more for it," Emmons noted. The traditional ways of building networks, using purpose-built hardware in networks engineered for peak traffic demand, don't work in this new paradigm.
"Through programmability and network awareness, we utilize that bandwidth more efficiently," Emmons notes. There is also the potential for further savings if it proves possible to use generic or even commodity hardware in the network, thus reducing capital costs. But that is more likely to happen in the upper layers of the network, and not at the optical layer, he said.
"Even if we can't get to commodity hardware, however, standards are important for interoperability and so we can present APIs so the network can be programmable," Emmons said.
Among the significant challenges Verizon is facing with its SDN plans is maintaining the five-nines reliability of telecom networks in this new SDN paradigm, once networks are more programmable.
"Today, we have well-defined processes to survive outages and maintain the high reliability that customers demand," Emmons said. "That is actually an impediment to the agility we are trying to accomplish. So the challenge is how to make these high reliability networks programmable in a way that is reliable and fault-tolerant."
Emmons outlined five areas -- network elements, network management, service orchestration, business model, dev-ops and network ops -- in which there are significant challenges or changes needed for SDN to deliver on its promise.
Starting with network elements, the move to separate hardware and software and to hopefully deploy more commodity hardware will increase the need for multi-layer awareness of what is happening in the network to maintain self-healing networks as bandwidth adjustments are made on the fly. Network management will need to be able to encompass analytics, alarms and more across multiple vendors and multiple silos, in order to track things such as inventory.
"Inventory is key -- not just knowing it is there but knowing it is correct, so that when you go to change something, you are changing the right thing," Emmons noted. "There is a lot of work to be done there -- I think that is a huge opportunity in the vendor community for figuring out how to get that inventory correct and maintain it" once programmability is in play.
Business models will move to more of a utility-based model, where customers can pay as they go and buy what they need. "The idea is the better it works, the more you buy," he commented.
Emmons supports the notion of moving to a dev-ops model for faster deployment of new software and services, but pointed out that most of the software for telecom networks comes from vendors. So there is the need for a "net-ops" model that lets network operators work more closely with their vendors to enable "faster, more frequent releases" that will let the operator deliver new features faster.
There are still core issues to be resolved around SDN, which is one reason why Verizon is working closely with its key vendors on its strategy, Emmons said. One such issue is a standardized approach to SDN controllers and another is determining how centralized control can be, given the potential latency involved in signaling traffic between controllers and network elements. In each case, a standard approach is needed, he said. And ultimately, hardware and software both need to be optimized to ensure reliability.
— Carol Wilson, Editor-at-Large, Light Reading