PALO ALTO, Calif. -- Carrier Network Virtualization -- NFV is moving beyond the basic blocking and tackling phase, but there are still mountains to climb.
While NFV is making its way past the proof-of-concept stage, there are still many challenges to address in order for it reach its full potential, according to Mike Heffner, vice president of product line management at ADVA Optical Networking . Heffner spelled out the NFV pain points during his opening keynote address Tuesday morning at the Carrier Network Virtualization conference.
ADVA's Mike Heffner talks about the challenges NFV faces during his opening keynote at the Carrier Network Virtualization conference in Palo Alto, California.
"So where are we today? My view is that we're kind of in the middle of this tug-of-war between the hype and the promise of what NFV can deliver and the reality around the challenges of what it means to implement NFV and do it well," Heffner said.
When ETSI laid out the fundamentals of NFV in late 2012, it seemed fairly straightforward, but Heffner said the reality we see today "is that NFV is complex."
"It's difficult to implement. It's difficult to manage and it's difficult to commercialize," he said. "There are challenges, equally large challenges, on the technical side that go beyond the bits and bytes and the blocking and tackling types of things. We really need to look at dealing with all of those kinds of things."
Today, Heffner said, NFV isn't about the basic aspects of the technology, but how it scales and rolls out across an end-to-end network, and how to manage it across that entire network.
The first key challenge on Heffner's to-do list was "openness."
"Most of the carriers I've talked to really want to avoid vendor lock-in, and so they are pushing very hard about having open, modular architectures," he said. "I can use different vendors' components within an entire end-to-end NFV architecture but in order to do that those different vendor components have to be able to talk over open standard interfaces using standard protocols."
The first challenge dovetailed into the second, which Heffner said was "fragmentation." For example, there are currently competing activities by open source groups that are defining different philosophies around what's going to happen at the MANO layer, Virtualized Infrastructure Manager (VIM) layer and SDN control layer, Heffner said.
"So while it's good that a lot of people are trying to push a lot of architectures and trying to move the ball forward, at some level we're splitting our brains a little bit because service providers have to place a bet on an architecture," he said. "Which architecture is going to win, which one do we want to go forward with? Or, in even a worse case, we're going to support multiple architectures. So you're just dividing resources and focus, and the whole ball is moving forward just a little more slowly.
"The vendors are in the same spot. We as vendors have a hard time picking one architecture."
Heffner said the industry as a whole was spending too much time working on multiple, competing architectures and philosophies, "but these things need to start coming together."
Another challenge area is VNFs, according to Heffner. The original goal of the ETSI white paper was that NFV was supposed to be completely automated and completely managed. It would not only instantiate VNFs but configure them as well.
"That's a great idea but in practice MANO can only go so far," Heffner said. "How do we move towards cloud-native architectures, decomposing VNFs into their sub-tenant functions, things like microservices architecture, and being able to control VNFs at a more granular layer?"
Manageability and operations also fell into the challenge category for NFV. Heffner said it was important to support these functions across hundreds of thousands of nodes in order to do mass software upgrades on infrastructures, on VNFs and for porting service changes from one cloud to another.
While security is always going to be important, the federation that is brought about by NFV in the different areas are exposed because virtualization opens up a number of new attack vectors that need to be secured and managed, according to Heffner.
The last challenge Heffner spoke about was "assurance."
"How do you give the quality to your customers? How do you do fault isolation, fault mitigation, root cause analysis when I've got new layers of infrastructure controlling packets?" Heffner asked. "I'm not just tapping on a wire now. I need to be able to monitor all the way up underneath a VM in a virtual switch.
"These are the next wave of challenges beyond the blocking and tackling that we hear about today. But the good thing is that the carriers are really starting to talk about these challenges. To me that's more indicative of buying decisions and really trying to think about what it's actually going to take to run one of these networks in the future."
Heffner also provided a shortlist of best practices moving forward. The first was not to try to "boil the ocean."
"It's good to have an architectural concept but you don't have to have the perfect architectural concept before you get started," he said. "That dovetails into a second comment. Don't let 'best' get in the way of 'good' or 'better.' There are incremental gains to be realized by starting small with NFV."
Other best NFV practices for carriers included starting with an easy use case or friendly customer in a controlled environment, and taking those lessons learned to improve and grow NFV incrementally.
— Mike Robuck, Editor, Telco Transformation