NFV is not—definitely not—living up to its touted potential, according to Masergy's Ray Watson.
During a panel discussion at Light Reading's NFV and Carrier SDN conference in Denver, Watson, vice president of global technology for Masergy, said that network function virtualization (NFV) hasn't met with the original "promises" that were laid out in the European Telecommunications Standards Institute's (ETSI) first NFV white paper six years ago.
"This is the fifth Light Reading NFV conference," Watson said. "ETSI wrote the NFV draft six years ago, and the promise that we were told around economics and agility simply haven't become a reality yet. We've been talking about how it would make sense to deploy commercial off-the-shelf software and hardware and put one VNF on it and it would make money, and the reality is that's just not the case. The commercial VNF vendors, the hardware vendors, et cetera, have not made it easy for us to realize that potential. I'm scolding some of the people in this room, but I'm actually not going to name names.
"Here are we at a critical point. We're halfway through what most people would consider the adoption of NFV, and simply we have not seen all of the promise that we were promised and all of the potential we were promised around agility and around economics."
When asked what needed to be virtualized in networks or how many virtual network functions were needed to make NFV fulfill its promise of lower opex, Watson said Masergy needed to deploy 2.2 commercial VNFs just to break even.
"We were promised it would be one [VNF] so that's only off by—we'll round it off, double," he said. "So if you can't deploy 2.2 VNFs, it's not even worth sending out the equipment at all."
Watson clarified that he was speaking about commercial VNFs and not open source VNFs, the latter of which "changes the math."
Speaking on the same panel, Antoine Clerget, chief technical officer of OneAccess for Ekinops, said that in regards to reduced opex, virtual customer premises equipment (vCPE) boxes were four times more expensive than the boxes that they are replacing, "but we've been optimizing the cost for 20 years; how do you expect it to be cheaper?"
Clerget said instead of one vendor selling a CPE function, now there are universal CPUs from vendors selling hardware, virtualized infrastructure from another vendor, and then VNFs from several vendors.
"How do you expect that to be cheaper?" Clerget asked. "We started [NFV] with the wrong expectations. I'm not saying in the long run that you will not get benefits from the new services and the benefits from opex reduction, but I think what we should focus on in the short run is bringing new user experiences. I think this is where we will start with the benefits. I think we started with the wrong expectations."
Clerget used the analogy of an Uber car versus a taxi. Both cars have roughly the same costs, but one is relatively cheaper to use and provides a better user experience for passengers.
"Your Uber analogy is perfect," Watson said. "You're right; we did go into this with the wrong expectations because we were told that it would be a no-brainer. It does bring immense benefits to the end user around agility as we move to DevOps where people are trying to move faster, easier and quicker. Sure, there's a great progress for NFV in that space."
Richard Piasentin, chief marketing and chief strategy officer for Accedian, said that SDN, NFV and 5G give service providers the opportunity to offer lower latency services at the edge of networks with end-to-end service level agreements that can't be matched by the likes of Google, Amazon Web Services and Microsoft. With telcos changing their central offices to data centers, they're also better able to deliver edge services.
The sunny side of NFV
In an interview at the same conference in Denver, CenturyLink's Bill Walker, senior director of strategy and advancement, said his company went into NFV with its eyes wide open.
"We knew it was going to be hard," he said. "It's a slow-moving legacy industry, so we had no illusions that it was going to be this dynamic earth-changing thing. I think we've been really smart about where we leverage it. We've had it in production and have been making revenue off of it since 2013. Those products are still out there.
"The promise from ETSI and the participants in ETSI were more about operational savings and converting the legacy infrastructure. We don't convert the legacy until we get our money out of it. So unless we see a viable need or business case to replace how we manage a core network, it's not going to be virtualized quickly."
Walker said that CenturyLink, which is on its third iteration of NFV, has focused on using NFV for new products and new revenue rather than focusing on operational changes.
"We think it's highly successful," Walker said. "We're angry at ourselves for not moving faster, but that's just a function of creating new products and new services. Doing the due diligence, the testing and reliability, the service assurance and the legacy IT work actually slowed it down. New products is where we're seeing the growth. We would rather focus on new products and new revenue than trying to focus on a operational change or a core network change for the sake of a core network change when we haven’t fully monetized the existing investment."