Network Functions Virtualization: Learnings from an NFV proof-of-concept

Next Generation Communications Blog

Network Functions Virtualization: Learnings from an NFV proof-of-concept

By: Asaf Peled, Cloud Solution Architect, Alcatel-Lucent's CloudBand activity 

We live in exciting times – again. I had been wondering if there was still room for fundamental innovation, for technological disruption – the worn out word. We had the Internet bubble that burst 2001/2002. We had the advent of the smartphone in the recent years. Great innovation, but somehow they reached a plateau. Faster processors, larger screens? It doesn’t seem to make a real difference. Apple’s engine showed the first signs of sputtering.

And yet, there is something brewing behind the scenes that makes the engineer’s heart beat faster: NFV and SDN, a bold new vision about the future of networks (read these blog posts about understanding Network Functions Virtualization and Software Defined Networks). Network functions are to be reduced to pure software - doing away with all the special purpose chips, circuit boards, and cabinets into which we have poured our brains to deliver the ultimate in features, performance, and reliability.

Telefonica is one of the original masterminds of the NFV idea, and a partner who truly shares with us the excitement about this new technology. I had the pleasure to work with a multi-national team dedicated to making NFV a reality. The objective of this proof of concept was to drive the vision of NFV and develop a blueprint for the future production of NFV infrastructure. We installed the first node of our CloudBand NFV Platform in one city, and then a few months later extended the installation with a second node in another major city hundreds of kilometers away. One of the interesting questions to explore was the remote management of a distributed NFV infrastructure. How quickly can we deploy cloud nodes in additional locations? Can we really operate the infrastructure without having to send engineers and operations staff around the country? How can we set up the network between multiple sites to react dynamically to changes to network function deployment?

 This latter question points to another major topic tackled in project: the relationship between NFV and SDN. Virtualizing the network, SDN should be a perfect addition to NFV as a platform to virtualize network functions. In fact, we deployed the Alcatel-Lucent CloudBand™ NFV Platform along with an SDN solution from Nuage Networks. This allowed us to push virtual network functions (vNF) policies into the Nuage Networks policy repository. The SDN solution then knew how data traffic from the VNF should be handled not only in a single data center but automatically extending across multiple data centers.

For me, the project was a milestone on the way to our NFV vision. It was the first project covering more than one site. It was the first project with SDN integration. It was an excellent opportunity to discuss with our friends and colleagues at Telefonica many of the questions that arise only as we go from the theory of the architectures developed at ETSI to the real networks, even if they are not yet serving real customers. One challenge was the deployment of the proof-of-concept alongside a live network. This meant we had to work within a very narrow maintenance window. And not only that: we had to undergo all the qualification and certification stages of a standard appliance-based solution being introduced into the network. At first, we were surprised of this requirement as NFV was invented to cut down exactly on these operational processes. But we quickly understood the situation. Once the NFV infrastructure was qualified and certified, any virtual network functions running on it no longer would have to go through the same process. There may still be a qualification and certification process for virtual network functions, but this process can be more automated and doesn’t have to repeat some of the steps already done for the NFV infrastructure.

Another experience from the project relates to the culture of collaboration and debate. If we want to be successful in NFV we must be disruptive and challenge each other and challenge the overcome ways. Maybe we, Alcatel-Lucent, were a tougher partner for Telefonica than others on this front. In one case, the Telefonica operations team came to us and requested user-configurable VM sizes (flavors).  We did not immediately comply with this requirement, but put up the challenge: does it make sense to have so many sizes? What is the impact on efficiency of resource usage? Together we concluded that we must adopt web-scale MOPs (Method of Procedure) to be really efficient and lean in operations. Much like for this operations team, we need to be there for the service providers. Together think differently. Collaboration is the word in NFV.

I am convinced I will be able to witness how virtualization and cloud will lead to massive changes in the telco industry. I am glad to be able to contribute to these changes. Could there be any better place to be?

Related Articles to 'Network Functions Virtualization: Learnings from an NFV proof-of-concept'

Featured Events