With recent initiatives like AT&T (with ECOMP) and Verizon with its new architecture whitepaper, operators have stepped up to describe in detail how they plan to build the network of the future. Central to both their models are two facts -- they want to wrestle control of networking from equipment vendors, and they don't think evolving SDN and NFV standards will do enough or be fast enough to meet their goals. Is this a true changing of the guard or just another phase? Here are some reasons to take these projects seriously.
Looking to learn more about NFV and SDN? Register to attend Light Reading’s upcoming NFV & Carrier SDN: SDN Integration in the Virtualized Realm taking place September 13th &14th in Denver. You can register now.
They take a top-down approach to the next-gen network. Both SDN and NFV were started at the bottom, with a technical approach that might or might not connect with any of the benefits needed to justify a real network change.
AT&T Inc. (NYSE: T),
Verizon Communications Inc. (NYSE: VZ) and operators in Europe like Telefónica and
Orange Business Services , have taken care to start with the benefits and work down to the implementation. That alone gives them a better chance of deployment.
All these operators have engaged next-gen network benefits by broadening their architecture from the narrow scope of SDN/NFV specifications. Their architectures prove that either SDN or NFV must be considered in a full operations and service context if they're to secure maximum capex and opex savings and realize new revenue opportunities.
They define the future as an evolution from the present. Network infrastructure has a long depreciation cycle, and that means few operators could introduce SDN or NFV all at once. However, limiting their deployment scope would also limit the benefits they can realize, and if early deployments are too small and too service-specific, they may not prove out either technology on a large scale. All operators have strategies to deliver some operations and agility benefits even for services built on current infrastructure.
The trend is to define "end-to-end orchestration" (EEO) as a function that sits above SDN, NFV and legacy management systems, and to use EEO for high-level service automation. Not only would that let operators include legacy elements in SDN/NFV services, it would allow them to automate services with no SDN/NFV at all. By creating service automation benefits even before a single SDN or NFV element is deployed, this approach can help fund early SDN/NFV deployments and reduce risks by giving operators an opportunity to become familiar with SDN/NFV-friendly operations practices.
They define an open-mandated, open source-facilitating architecture. All of the operator architectures presented so far introduce multi-layer service models, where each layer can be an independent implementation driven by its own modeling language. The models that drive these implementations are increasingly based on the cloud standard TOSCA (Topology and Orchestration Specification for Cloud Applications) at the higher levels, and on the YANG device/connection model below. Operators can take advantage of any implementation based on these models, including open source implementations.
Operators prefer an open source approach to SDN and NFV, but none have so far been able to identify a complete set of open source tools to implement their approach. By mandating an open, layered, model-driven architecture they can incorporate proprietary elements if there are no suitable open source tools yet available or use a proprietary component if it offers better capability, without compromising their multi-vendor mandate.
If these operator architectures meet their goals, then connection technology -- Level 2 and 3 devices -- is headed for commoditization. Vendors won't like that very much, but there's some better news hidden in the details.
First, obviously, all the operator visions of the future increase the role of fiber in general, and agile optics in particular. Many operators tell me they expect their spending on fiber and ROADMs to increase annually well into the future. This is good news for optical vendors, but most network vendors have at least a foot in the optical-core business, and the New IP will be more and more optics-driven.
The second point is that mass-market IP service -- the Internet -- isn't seen as a specific target for virtualization via either SDN or NFV. At least some router technology likely will be in place, even if it's just a big router at the boundary between access/metro networks and the core.
The third point is that white boxes are going to look "gray" for a long time, despite operator goals to the contrary. Because of the long depreciation cycles for network gear, operators will have to evolve their physical infrastructure even if they can create an orchestration revolution above it. That will give vendors a chance to introduce things like SDN/OpenFlow in current products to supplement traditional adaptive routing or switching. Because white-box switches will be introduced, at least initially, into current networks, the vendors may develop special features to support coexistence. If those features are valuable enough, operators would still probably buy the products as long as their basic features still fit into an open model.
Everyone agrees operators are now focusing on the business case for network evolution, not just the technology options. Given that, vendors must focus on how their customers' goals for benefit creation can be met without destroying the vendors' business models in turn. That means accepting change, even embracing it, because operators are showing us that the New IP won't be built by hunkering down on legacy technology.
— Tom Nolle, President, Founder, and Principal Analyst, CIMI Corp.. Follow him on Twitter @CIMICorp. Special to The New IP