Answering the NFV Management Challenge
Amidst all the hype around software-defined networking (SDN) and network functions virtualization (NFV), three realities have emerged: First, these two technologies are essential to the future success of telecom networks, because they hold the best hope of reducing costs and adding new revenue-generating services. Second, the changes they represent are profound, touching every part of the network, and so they won't happen quickly.
The third reality follows on from those first two and may be the most important: Telecom service providers have no choice about adopting SDN and NFV, so they'd better be about the business of doing that now.
Major network operators clearly are aware of these realities and are pressing both their vendors and their associations to be moving ahead with greater speed than usual in making plans for this transition. Vendors report this activity, without naming names, and the service providers themselves are publicly talking about the urgency of this action.
So after all our talk of the challenges of NFV and SDN deployment, we're going to focus on the solutions, and there are quite a few. This article will focus on CloudNFV's efforts, with the next stage being a look at the vendor-specific projects. CloudNFV is a group of vendors that came together around the thinking that virtualizing functions is not enough -- it is essential that service providers move to a private cloud environment for managing the widely distributed functions of an true NFV deployment. (See New Group Ties NFV to the Cloud and NFV in the Cloud: It's Complicated.)
As industry analyst Tom Nolle, the godfather of CloudNFV, explains in his recent CIMI Corp. blog, the organization has a different take on the Management and Orchestration (MANO) layer of the European Telecommunications Standards Institute (ETSI) NFV group and has created a set of proof of concept proposals for that organization as well. The first of those is being delivered this month.
CloudNFV is trying to look at both management and orchestration as applications that can run off a unified data model that is very agile, Nolle says. While using a single data model for simplicity's sake, the organization describes the data model in two levels -- Active Contract and Active Resource.
"Active Contract is the portion of the data model that relates to service orders, service templates that you would create an order from, and a description of components -- in other words, it is kind of the service side of the picture," Nolle explains. "Active resource represents a data model that describes the state of the infrastructure."
In this model, the NFV's MANO function is an app that connects Active Contract to Active Resource, he adds.
"We call the management process derived operations, because when we describe how a service is created from virtual components, we also describe how management of a given set of components is derived from Active Resource, with what we call a visualizer," Nolle says.
So if a Virtual Private Network service uses a combination of real network resources -- a router, for example -- and functions that have been virtualized, such as a firewall or a load balancer, service assurance and status reports on that service are delivered from a combination of the real network management of the router and the visualizers for the virtual functions.
That is, of course, a very simplified explanation of a more complex problem, but it begins to show how CloudNFV sees the transition to fully virtualized network services. There is a visualizer application that corresponds to the virtualized function and can derive its status and describe it using the existing notions of customer-facing services and resource-facing services as described by the TM Forum's Frameworx and therefore modeled into a TMForum-compliant Operations and Support Systems/Business Support Systems process -- if that's the choice the service provider makes, says Nolle.
"One of the problems is that there isn't agreement, even within the service providers, as to whether they should integrate into their existing OSS/BSS systems or try to do something totally different," Nolle says. He describes conversations with executives from the same company whose views on handling legacy support systems are 180 degrees apart.
The CloudNFV approach finds something of a middle ground, where virtualized things are managed in a more virtual way but existing network resources and existing operations can still function as they have, where that makes sense, without being blown up in the process. The worst-case scenario, Nolle adds, would be to force old management techniques onto new virtualized functions and the new services they support.
"This is what I mean when I talk about selling virtualization short," he says. "You limit the power of virtualization by replicating the old process, by managing things as if they are real."
There is much more detail about how CloudNFV will do this on its website and much more detail to come this fall, when the participating vendors will go public and the first demo, an IMS deployment, will be publicly staged.
— Carol Wilson, Editor-at-Large, Light Reading