(Above: Rob Lloyd (left) and Surya Panditi formally unveil the NCS line.)
One could argue, though, that BSkyB, the system’s first announced customer, is using the box as just a big router. The NCS’s other features — optical/IP convergence, and the trappings of software-defined networking (SDN) — are intriguing but won’t come into play until later, if at all, said Damon Pegg, Sky’s chief network architect.
Pegg was talking to analysts and press via telepresence late Tuesday morning (pub time, in the U.K.) as part of an extended briefing accompanying the NCS launch. During that time, officials described the Cisco NCS products, including the smaller NCS 4000 router and the NCS 2000 optical-transport box, as working in conjunction to form a fabric or “supercore” for the service-provider network.
That’s different from the usual network hierarchy, where really big routers sit in the core and smaller routers sit at the edge. Cisco pictures the larger NCS 6000 in particular, going into some metro and edge deployments. And of course, the NCS core is intended to be programmable, with connections set up and torn down in a more automated fashion.
Sky is interested in that big picture, and that’s why the carrier chose the NCS over options including Cisco’s CRS-X, another core router announced earlier this year. “We’ll scale horizontally, in a way CRS did not,” Pegg said.
But the overriding concern was the relentless climb of bandwidth demand. The NCS 6000’s first job at Sky will be to provide 100-Gb/s connections across the network core.
Sky has taken delivery of an NCS chassis but won’t be installing any until January. A second phase of deployments is planned for March or April. By late summer, Sky expects almost all its core traffic to be traveling over NCSs.
“For us, it’s effectively a replacement for what we have in CRS,” providing headroom for 10 years of bandwidth growth and a significantly lower electricity bill than CRSs would incur, he said.
In looking for an upgrade beyond Cisco’s CRS-3 core routers, Sky executives started with a blank slate — no assumption that they had to stick with Cisco. So, software-defined networking (SDN) features and the convergence with the optical layer did matter. Candidates included some “left-field” ideas out of the optical industry, which Pegg said were interesting but ultimately “lacking in maturity.”
Not an Overnight Job
Cisco is saying the NCS family’s real purpose is to be the fabric for Cisco routers to plug into, all orchestrated by Cisco Quantum (which comes from acquired startup Cariden). Cisco also claims this core could play nicely in a multivendor environment, as other boxes can connect to the NCSs through Cisco’s onePK set of APIs.
Building that new network is going to take a while, though, even when it comes to the less radical concepts such as the NCS’s converged control over the packet and optical layers.
“We’re going to see carriers, over time, embrace this degree of convergence. It’s not going to happen overnight,” said Bill Gartner, vice president of Cicso’s high-end routing and optical business. “What we’re effectively doing is giving them a knob to turn,” letting them design the network without having to predict which kinds of traffic will be running on it. Theoretically, the network could adjust to, say, a changing balance between Ethernet and Optical Transport Network (OTN) traffic by creating the right kinds of connections.
The plan would seem to stamp a lame-duck status on the CRS-X, considering the NCS box seems like a futureproofed version of the same thing. It’s bigger, with theoretical capacities in the petabit-per-second range, has processing power to support network functions virtualization (NFV), and fits better into an SDN future.
But Cisco says it had to update the CRS line for the sake of the installed base. Multichassis CRS deployments (and Cisco keeps stressing that it really does have these) needed a way to increase capacity by adding a much larger, compatible box. The CRS-X also give service providers an upgrade to 400-Gb/s slots without having to divorce from the CRS architecture.
Getting Holistic With Apps
The programmability of the NCS line might take a while to blossom as well. It could be harder to do than it sounds, said Paul Parker-Johnson, an analyst with ACG Research.
“Say you have somebody in your application partner ecosystem who wants to do an Internet of Things app, and you want to stage it at different points throughout the infrastructure. How do you develop that app?” Parker-Johnson said. “If it’s supposed to be deployed holistically, is it going to be developed holistically? There’s an assumption of deployability which, if it’s there, is Nirvana.”
It’s also possible to offload part of the control plane onto blade servers (the ones in Cisco’s own UCS, specifically), but Parker-Johnson had a question about that, too. “From an orchestration point of view, does that stuff behave consistently with what’s left on the metal?” he asked.
Parker-Johnson believes in the idea behind the NCS. It’s just that any new way of doing things will inevitably hit speed bumps.
Read more breaking news on SDxCentral: