Jim Metzler recently penned an excellent conversation-starting article asking, "Does NFV have a place in the enterprise?" He notes, among other arguments,

"I don’t know of any enterprise that currently has, or that ever will, virtualize all of the functionality that ETSI is focused on because a lot of that functionality just isn’t applicable to an enterprise network. However, the majority of enterprises have virtualized at least some functionality."

I think it’s important to start with noting the abbreviation NFV has three discrete components: Network Functions Virtualization.

Interestingly enough, service providers tend to fixate on the network functions that will be virtualized and later orchestrated in one glorious automated workflow. From a service provider perspective this makes sense; their entire SGi network is laden with boxen designed to get traffic from their mobile network to the Internet while generating revenue with value added services functions. virtualizing those functions makes good economic and operational sense. But their end goal is to move traffic. Traffic that is broken down into what it is: video, text, unified communications but ultimately ignores the application itself. YouTube or Vimeo isn't really all that important; what's important is that the traffic is video and subscribers expect all video to play seamlessly on their latest iGizmo.

Enterprises, on the other hand, tend to fixate on the form, that is, network virtualization and the characteristic agility and cost savings that can bring.  Their primary goal is the application experience; performance and security of applications are paramount, and to do that increasingly means enabling a per-application service model that cannot be economically sustained without embracing virtualization, at least for those services tied to specific applications. So they focus on virtualization because they need to capitalize on the ability to leverage COTS hardware, reduce the time to procure and provision, and service-enable IT using automation and orchestration to reduce the operational impact of the increasing number of services in their infrastructure.  

What that means is there will be an NFV for the enterprise but it's unlikely it will be "the" NFV as envisioned and implemented by service providers. I still maintain that NSV (Network Service Virtualization) is the likely form of NFV in the enterprise, as it aligns more closely with the enterprises' tendency to focus on form over function.

Both approaches are equally valid, mind you. But an enterprise is not a service provider and vice versa, thus leading to variations on a theme, if you will. If you recall earlier this century, IMS was the dominant architecture in service provider networks. IMS was, at its core, SOA - but with very carrier-focused perspective. The two were not "interchangeable" even though they were based on the same core principles.

So we're likely to see an enterprise variant of NFV (I'm rooting for NSV, but you knew that) but like children are shallow copies* of their parents, so too will enterprise NFV be of its service provider parent.


* I'm using the term in the programmatic sense, somewhat like differentiating between passing parameters by value versus by reference. This isn't helping, is it? Trust me, it’s not bad.