Cisco is announcing what’s essentially its next Huge Freakin’ Router, bringing software-defined networking (SDN) and network functions virtualization (NFV) into its service-provider portfolio for the sake of network programmability.
The Network Covergence System (NCS) is a set of systems aimed at the service-provider network — so don’t get all excited about tying it to Insieme, the spin-in that’s developing a huge switch (or something) for the data center. Insieme’s launch is still pending, probably until November, it now appears.
The NCS line, being introduced Tuesday, includes a few products:
- The NCS 6000 is the aforementioned core router, shipping today with available 1 Tb/s cards (ten 100Gb/s interfaces) and an eventual capacity of 5 Tb/s per slot, Cisco claims. (Light Reading broke the story about this system in June, based on a document that called it the CTR.)
- The NCS 4000, a smaller box for positions closer to the network edge; it supports 400 Gb/s per slot. This one’s not shipping until the first half of 2014.
- The NCS 2000, which isn’t a router at all but is part of the optical network (a reconfigurable optical add-drop multiplexer (ROADM), if you have to know). It’s shipping today.
Cisco is using the NCS 6000 and 4000 as the vanguard in the company’s strategy for SDN and NFV in the carrier network. You can think of it as Cisco extending the fabric beyond the data center, something that large carriers are interested in seeing, says Ray Mota, an analyst with ACG Research.
The ‘SDN’ Pieces of NCS
Cisco isn’t emphasizing the SDN aspects of Tuesday’s launch, instead putting the Internet of Everything (or Internet of Things, as most other companies seem to call it) in the headlines. But the NCS touches on a few SDN/NFV bullet points — or functions closely related to SDN, anyway.
1. Fast provisioning and moveable services. Cisco had previously converged the optical, routing, and switching control planes into one set of software called nLight. The software can redirect all those types of workloads around the network as necessary.
Now Cisco is also talking about nLight on the NCS being able to provision overlays across the IP and optical layers simultaneously. That capability will start just at Layer 3, being able to move workloads between access, edge, and core networks arbitrarily. What it should add up to is the ability to quickly put up a service where it’s needed.
2. Scaling (and separating) the control plane. The nPower X1 network processor, announced earlier in September, powers the NCS routers and gives them the option of using large amounts of off-chip memory as necessary. This means services running on the chip can tap a memory pool that might as well be infinite, Cisco says.
More specifically, nPower can offload the control plane as necessary, moving it to server blades in Cisco’s Unified Computing System (UCS). Cisco sees this as a scaling tool, a way to expand the control plane as in-system resources get tapped out.
Cisco notes that the CRS-X core router, the NCS 6000’s predecessor, is using the nPower X1 as well. So, the CRS-X can offload its control plane as well. “This new paradigm is making its way into the network, and the NCS is the lead dog,” says Stephen Liu, Cisco’s director of service provider marketing.
3. Splitting the OS. The NCS systems can run multiple versions of the IOS XR operating system, partitioning the system to keep different instances isolated.
It’s similar to the concept of splitting a router into multiple virtual routers, but it allows for the routers to be even more different from one another. Liu used the example of running a wholesale service and a VPN service from one router. Two versions of IOS XR could get loaded, and the router ports could get divided between the two types of services.
4. Distributed NFV: Placing services anywhere. More specifically, this is about moving services from one NCS box into another as necessary. This could be handy for creating in-router service chaining, by moving a group of services, along with the necessary policy and state information, into the same box.
That would avoid the problem of tromboning — having traffic repeatedly move out of the router to get to one service after another. “The location is still a cost I don’t think people have analyzed yet,” Liu says.
Cisco’s claim is that Layer 3 is the right vantage point for stitching together the optical, switching, routing, and applications layers, with policy included in-line. In other words, the convergence of Layers 1 through 7, in Cisco’s mind, starts at Layer 3 — not a surprise, given Cisco’s longtime standing as a Layer 3 kingpin.
The hosting and portability of services is a concept that won’t be limited to the NCS. Cisco officials hint that the CRS core router and ASR edge router lines could participate here as well. “It is an IOS XR strategy,” Liu says. “You’re going to see this make itself relevant in other parts of the network.
Cisco needs multiple platforms involved here, because functions from different environments — edge vs. core, for instance — will get virtualized. “Cisco wants to make sure that its platforms have that ability to run the virtualization inside,” Mota says.
Evolving Beyond Next-Gen IP
As you’d expect, the strategy involves lots of Cisco gear, covering multiple parts of the network (edge and core) and multiple layers (optical, routing, orchestration, and some applications). While the NCS and resulting fabric are open, Cisco is of course pointing out that Cisco pieces can complete the network — namely, Cisco Prime and Cisco Quantum network management.
Cisco calls the resulting network the Evolved Programmable Network. You can consider it the next life phase for what the company has been calling the IP Next-Generation Network (NGN).
The important thing is for the router to “adapt topologies on the fly, change them for different service-level agreements on the fly, and allow services to be programmed by the users,” says Len LuPriore, a Cisco senior manager of strategic marketing. “The service provider is able to let the customer DJ their own solution. That’s the crux of what we’re going to.”
Well, almost. It will take time for end customers to start programming their own services, even if the nPower X1 is programmed in C. At first, Cisco expects its tier 1 service providers to do the programming; they’re the ones with programming teams available for the task. One possible benefit for service providers would be the ability to create new features without waiting for vendors to develop them, LuPriore says — and yes, he’s aware that he works for a vendor.
Cisco is not yet announcing any ready-made applications to run on the nPower X1, nor is it announcing any programming services to help service providers create applications.
Join SDNCentral’s DemoFriday™ on Sept. 27 to see Cisco demonstrate its Extensible Network Controller (XNC). The event has ended, watch full presentation here.
Read more breaking news on SDNCentral: