Last week I was at Upperside Conference’s 20th annual MPLS +SDN+NFV World Congress in Paris, France, along with a record 1,500-plus attendees. What struck me were the number of loyal attendees, some of whom have been attending this gathering for 20 years. The content covered the gamut from software-defined wide area networking (SD-WAN) to MPLS, segment routing, IPv6, security, and blockchain. Hot topics like artificial intelligence (AI) and machine learning (ML) were covered at a co-located conference, called parallel ai.net, also hosted by Upperside.
During the Upperside Conference, I moderated a panel session with panelists from Telefonica, Orange, RAD, Nokia, Colt and Google. The panelists were candid about both the positives and negatives of software-defined networking (SDN) and network functions virtualization (NFV) deployment. And while all of them believe that SDN and NFV hold great promise for the telecommunications industry, they also admitted that there are still some challenges to resolve before operators realize the full benefits of SDN and NFV.
RAD Data Communications CTO Yaakov Stein brought up how Bruce Davie (one of the key technologists from Nicira, the first SDN company to be acquired in 2014) characterized it as a bubble: “If No. 1, everyone defines it the way they want to. No. 2, the definition gets larger and larger until anything you do is that and No. 3, when people get up on the stage and say, ‘I was doing SDN before they called it SDN, that means it’s a bubble.’ With that definition, my view is that SDN qualifies as a bubble, as do other related technologies, like SD-WAN, for instance.”
In addition, the panelists pointed out that many of the principles of SDN — such as the separation of the control and the data plane — were already present in networking before SDN. And as Stephane Litkowski, network architect with Orange Business Services noted, a lot of SDN principles are borrowed from past networking innovations. Centralization of control, for instance, is something we’ve had for some time.
SDN – The “Kick-in-the-Butt” We Needed
And while most of us in the industry recognize that there is hype in SDN, Wim Henderickx, director of network consulting engineering for the IP division at Nokia, noted that the hype was the “kick-in-the-butt” that the networking industry needed to drive innovation. And Vijoy Pandey, head of engineering (data center, cloud, backbone) at Google, pointed out that SDN has been implemented successfully in many places. For instance, Google’s intent-based networking is built on successful SDN foundations. Litkowski added that intent on top of SDN is where the big change is.
In general, the panel recognized that while SDN was a bit hyped, it has created some real-world innovation. After all, the concepts of SDN are now present across all parts of the network with programmability, centralized control, and openness in network platforms across the data center, campus, and wide area networks.
NFV – Both Hype and Reality
Moving on to NFV, Jimmy Ehrbar, senior manager of architecture and design at Colt, said that NFV was already a “kind-of” reality, whether we subscribed to the hype of it or not. While we’re certainly not near the end of the journey, we are at a point where we do see NFV working, especially in mobile. Though he acknowledged that there’s a long way to go to mature the overall platform, and vendors have to step up to make NFV more of a reality.
Stein pointed out that for some applications of NFV, such as packet forwarding at the edge, capex per processed GB might be higher with white box servers and NFV than with proprietary purpose-built hardware. This cost increase could be offset by reductions in opex and use of hardware accelerators. And the panel agreed with Stein that additional hardware assist would be necessary in some NFV deployments to make it cost-efficient. As a counterpoint, Pandey said that Google was using NFV, and that sometimes it wasn’t purely about cost, but about flexibility and agility. And that we’re still working through what should be disaggregated and de-composed into software functions.
The panelists also acknowledged that today’s NFV deployments still face numerous challenges including interoperability and lack of standardization. And Diego Lopez, the chair of ETSI NFV ISG and a senior technology expert at Telefonica, suggested that perhaps we rushed to commercialize NFV (and SDN) before we laid down the proper foundations and principles for the technology.
Bottom line: SDN and NFV are both hype and reality. Certainly, there’s been a lot of buzz about both, but at the same time, as Google’s Pandey pointed out, if SDN and NFV were just bubbles, they would have burst by now. It’s because there’s a real need for both of these technologies that the industry has kept investing in them.
SDN and NFV – Make Our Own Reality
The march toward SDN and NFV won’t stop. Litkowski said that NFV is a reality today, and he provided one of the key observations from the panel: “NFV is a reality today, but is it the reality that we wanted?”
The panelists agreed that there are enough real-world proof of concepts (POCs) and limited production today with specific use cases that demonstrate value. However, timeframes will need to be more realistic, and there are still numerous problems left to be solved. But as the panel concluded, it’s up to all of us in the community – vendors, service providers, cloud service providers, and enterprises alike to keep going at it but be more realistic about expectations we set and be prepared to candidly discuss the real-world challenges we face and work hard to solve the challenges one at a time. It’s up to us to make NFV and SDN the reality that we want.