Network functions virtualization (NFV) might still have its complications, but the “good” of its deployments shouldn't be forgotten among the “bad” and “ugly” factors that are still being worked on.

That was the theme of “NFV: The Good, the Bad, and the Ugly,” an Oracle OpenWorld session led by Douglas Tait, director of global markets.

It was the sequel to a similar session Tait hosted last year, the goal is to provide an overview of the state of NFV. As with last year's session, panelists pointed out NFV's issues but left an optimistic note by summarizing NFV's progress. This year's discussion was framed by bullet points from panelists from Intel and Orange.

A couple of bad/ugly issues from last year's session came back. For example, there's still concern that too many open source consortiums might be working on NFV.

The hidden “good” there is that open source took over the NFV process from the start, with teams producing code instead of waiting for standards, said Christian Buerger, a member of the software-defined networking (SDN)/NFV marketing team at Intel. But he admitted the situation is still “reasonably ugly for somebody that wants to buy today.”

That was a recurring theme during the discussion. NFV has shown signs of progress during the last 12 months, and there's reason to be optimistic about the whole picture coming together. But for someone hoping to tackle a large-scale, multivendor NFV implementation right now, some pieces are still missing.

For example, Wojciech Beśka, director of IT architecture for Orange, listed a lack of key performance indicators (KPIs) as an “Ugly.” His team at Orange Polska in Poland has worked on several NFV proofs-of-concept (PoCs). He said his team would have liked to turn to standardized KPIs to evaluate the health of a full-service chain.

NFV also isn't available as a single “product,” a point Tait had touched on last year. As with OpenStack, that leaves somebody with the job of assembling pieces. That could be an impediment to NFV succeeding, Buerger said. “Providing the Lego blocks will only make one set of ecosystem partners happy, and that's the systems integrators.”

In other ugly points, Beśka believed that a 100 percent NFV architecture “still seems to be a dream,” and Tait pointed out that security hasn't been as much of the NFV discussion as it could be.

So, what's good about NFV? The group found a few things to be happy about that have emerged in the past year:

  • Use cases for NFV are well understood, following the spate of PoCs that emerged. New technologies don't always have use cases in the bag.
  • NFV management and orchestration (MANO) is beginning to mature. There might be too many MANO efforts around, but at least the problem is being addressed.
  • Virtualized software performance is looking good. In NFV's early days, there was some concern that this might be a problem.
  • NFV gets the network and IT groups to work together, and both benefit from one another's requirements.

Overall, panelists agreed that NFV has matured. Customers have gotten to the “now what” phase, Buerger said. They've implemented NFV in some minor form and are now looking at the long term — not just in terms of network architecture, but also in areas such as training.

And implementing NFV “doesn't necessarily take a rewriting of the entire code,” Beśka said. Orange managed to insert current applications directly into the NFV infrastructure by using scripts.

Rewriting the code is probably inevitable for someone who wants the full benefits of NFV, he added. But his point was that a carrier can get the first taste of NFV without having to rewrite applications.