With all of the talk about network functions virtualization (NFV), we know that service providers are interested in transforming themselves by moving to a cloud model. But does it require wholesale change? Or can service providers move there incrementally?

The feedback I get from service providers indicates that they know it’s not possible to virtualize the entire network at once, but they can take steps to improve operations, improve automation, and reduce costs. Each service provider or network operator will take a different approach to NFV evolution.

“While many of the building blocks like ETSI NFVi and distributed SDN control will be the same across carriers, the implementation models will be unique to each carrier,” wrote Michael Kozlowski, vice president of product management at Windstream, in response to an email I sent to some carriers with questions about their NFV evolution. “Service models will be individual, and written languages like TOSCA and YANG. The cultural adaptations that each company takes will be individual based on their specific business model and customer base.”

NFV and its cousin, software-defined networking (SDN), are described as new platforms that can replace the old system. But the reality is service providers, as opposed to many hyperscale cloud providers, have a larger installed base of legacy equipment and services, as I’ve pointed out in this series of articles on NFV. This means they’ll take more incremental steps in moving to the cloud platform.

John Donovan, head of AT&Ts network, has described the service provider’s move to NFV as a series of phases. In the first phase, the company virtualizes some of its most important packet network functions, such as the mobile packet core and security functions. The second phase is quite bigger – as it evolves a larger scale disaggregation of hardware from software to deliver services from a cloud platform.

Yet another component of AT&T’s move to virtualization is the development of Enhanced Control, Orchestration, Management, and Policy (ECOMP), an orchestration platform that integrates with its operation support systems (OSSs), which it plans to release as open source. Over time, AT&T has said it wants 50 percent of its software to be open source, up from 5 percent more recently.

Most of the operators we’ve contacted speak of a step-by-step evolution, not a wholesale rip-and-replace.

When I asked Telefónica about its NFV evolution, Javier Gavilán, planning and technology director and global CTO, said in an email that each segment of the network requires planning and analysis.

“We are always looking for improvement in our networks,” said Gavilán. “NFV is an enabler to do that. Of course we think that NFV technologies can give us a lot of advantages, but we always advance our deployments using TCO-driven analysis and following a pragmatic approach.”

The conclusion we can reach is that it is going to take years to build a fully operational NFV stack in the service provider market. If there are themes developing, it’s that we now have strong offerings for NFV infrastructure, virtual network functions (VNF), and SDN, but there is more work to do at the orchestration level of the stack, especially as it relates to tying into the OSS. This is a process we have described as “lifecycle service orchestration.”