NFV and SDN: Paving the Way to a Software-Based Networking Future

Don Clarke
Principal Architect, Network Technologies

Mar 23, 2015

When ONF Executive Director Dan Pitt invited me to contribute a blog post, it brought to mind our interaction in the summer of 2012 on how to treat SDN in the seminal NFV White Paper I was then editing. The operator co-authors were keen to ensure that SDN and NFV were positioned in the paper as being complementary. This was important because we wanted to create momentum for NFV by highlighting use cases that did not require the then perceived complexity of SDN. As soon as the ETSI NFV Industry Specification Group (NFV ISG) was launched, we engaged with ONF, recognizing its key role in championing an open SDN ecosystem. And in 2014 the NFV ISG entered into an MoU with ONF to facilitate joint-work.

The vision for NFV was compelling because the benefits could be readily attained. By replacing network appliances based on proprietary hardware with virtualized network functions (VNFs) running on industry standard servers, operators could greatly accelerate time to market for new services, and operations streamlined through automation. Moreover, important NFV use cases (e.g. virtualized CPE) would not require massive systems upgrades — a huge barrier for innovation in telecoms. We are seeing this first-hand at CableLabs, where we have been able to prototype virtualized CPE for business services and home networks on a two-month development cycle.

In contrast, the simplified definition of SDN- the separation of control plane from data plane -in my mind does not adequately convey the compelling benefits of SDN. The term ‘Software Defined Networking’ should mean just that, every element of the network, including the VNFs and network control should be implemented within a fully programmable software environment, exposing open interfaces and leveraging the open source community. This is the only way to create an open ecosystem and to unleash a new and unprecedented wave of innovation in every aspect of networking.

NFV releases network functions “trapped inside hardware” (a description I stole from an HP colleague) achieving tremendous benefits. But VNFs must be dynamically configured and connected at scale to deliver tangible value. While today’s telecommunications operations support systems (OSS) are adequate for static NFV use cases, the real potential for NFV to transform networking can only be realized through SDN control. Consequently, SDN represents much more than the mere separation of control plane and data plane.

Given telecommunications networks are deployed at massive geographic scale, it is a hard sell to convince thousands, or even millions of customers that their services will be migrating to a new network platform where their services will not be quite the same but prices won’t go down. Couple that with the significant time and cost to upgrade the OSS, wide ranging operational processes changes, and the need to validate that the new platforms are sufficiently stable and reliable, not to mention the obligations of regulation, it is not surprising that there is hesitancy to contemplate significant telecommunications network transformations.

Consequently, the telecoms industry has resorted to decades of incremental network upgrades which have piled legacy functionality on top of legacy functionality to avoid the costs and risks of wholesale network and services migration. In the face of these realities, SDN was perceived to offer insufficient benefit to justify significant investment except in niche areas where it could be overlaid on top of existing systems. Furthermore, the idea of logically centralized SDN control is very scary to network designers who don’t readily understand abstract software concepts and who lose sleep striving to deliver reliable connectivity at massive scale, with relentless downward pressure on costs.

Just over two years into the NFV revolution, it is clear that the emergence of NFV has galvanized the industry to embrace software-based networking; short-circuiting a transition that might otherwise have taken years. The revelation that NFV can be deployed in digestible chunks, without massive system upgrades has forced network designers to take notice. After all, it is difficult to ignore a pervasive industry trend when vendors’ product plans have morphed into software roadmaps!

Given that NFV is now accepted by all major network operators and some have already made significant announcements, there is no turning back. Leading vendors have committed to NFV roadmaps and analysts talk about ‘when’ and not ‘if’ NFV will be deployed. More importantly, SDN and NFV are now frequently discussed in the same breath. In my mind, the distinction between NFV and SDN is becoming an artifact of history, and the terms will ultimately be subsumed by a software-based networking paradigm, which itself will emerge as an integral aspect of Cloud technology.

The emergence of NFV with SDN is accelerating the evolution of cloud technologies to satisfy the stringent requirements of software-based telecommunications networks. Whereas a web service could momentarily stall with minimal customer impact while a virtual machine reboots, some business-critical network services cannot tolerate loss of connectivity even for a few milliseconds. Therein lies both challenge and opportunity. Challenge because meeting stringent telecommunications availability and performance requirements is not easy as evidenced by the ETSI NFV ISG’s deliberations. Opportunity, because I foresee an unprecedented wave of telecommunications innovation on a par with the birth of the Internet.

Carrier-grade network resilience (e.g. 5-nines and beyond) will be achieved by pooling virtualized resources, fault management will be supplanted by autonomic self-healing networks that can not only withstand equipment failures but can even rapidly recover from large scale natural disasters by instantly migrating network capacity to remote location as demonstrated by NTT DOCOMO et al in the aftermath of the Fukushima disaster. And exciting new routing paradigms such as intent-based networking and content-based networking will become feasible in a much earlier timeframe with innovation galvanized by the potential for imminent experimentation on deployed infrastructures. I could go on…

The genie of software-based networking — where synergies between NFV and SDN result in significantly greater capability than either could deliver alone — is now truly out of the bottle. The ultimate challenge is to encourage growth of an open telecommunications ecosystem, where operators and vendors can work together to create and deliver value to their customers. Energized by the NFV ISG and ONF, among other industry groups, and open source projects that are becoming increasingly important, the reality is just around the corner.

Don Clarke is Principal Architect for Virtualization Technologies at CableLabs and Chairman of the ETSI NFV ISG Network Operator Council.