Hype is rising around software defined networking and network functions virtualization. But along with hype, there is also a lot of confusion. Some people are even swapping between these two technologies as if they’re the same. Both technologies are about to play a critical role in the next generation of telecommunications, data center and enterprise networking. However, both the definitions and scope of SDN and NFV are still evolving.

Adding to the confusion, the benefits of NFV and SDN are similar: cost efficiency, flexibility, ease of scalability and quicker service introduction. To make this even more perplexing, the terms come from two origins. The NFV terminology comes from the telco world, while SDN is from the IT domain. Part of the problem is that, so far, the telco world hasn’t seen much in the way of direct vendor positioning for NFV, despite the fact that many analysts and experts predict that it will likely have more of an impact on the network equipment market than SDN will.

What makes NFV and SDN possible? The capability of processors has finally reached a point where network functions can be deployed on commercial off-the-shelf hardware. The multi-core, virtualization and hypervisor technology enable multiple functions to occur on the same processor. Today’s processors are clearly capable. IP networks have enabled easier connectivity to network nodes and connectivity to network nodes that have been disaggregated, all at the speeds required. This allows separation of the control plane and data plane, for instance, enabling SDN in the first place, or separation of the application from the media server, making NFV possible.

Let’s examine the benefits of a software approach.

Capex reduction: NFV software would run on COTS hardware and when that has happened in other industries costs for that product set and solution typically come down. A related piece to this reasoning is savings on chassis. An NFV node can run on a single board computer and that goes into a common chassis with multiple other network nodes, also on single board computers. This results in savings on chassis as one wouldn’t need a chassis for each network node. There would also clearly be sparing savings since the COTS components are less expensive, and they would be more commonly available, obviating the need to as much on-site sparing. 

Opex reduction: A side benefit of managing NFV is if network functions as described above sit in a common side running on single board computers, then management of the nodes becomes easier. One could also make an argument that managing multiple vendors would be easier with SDN, especially if standards are in place for how the software interacts on the network. I have heard for instance talk of an SDN controller that would manage the apps interacting with the network functions. This would be able to configure the network and collect performance on the entire network. 

Service delivery: Another important consideration is that service velocity is increasing. New services are being rolled out quicker. Agility with respect to the network elements are required to be able to meet these needs. Software-based network elements enable this agility because of ease of scalability and ease of adding newer functions such as transcoding. And what about needs based services? That is, there might be a spike of traffic or service and you need more service capability instantly? With a software-based network element, adding more capacity can be enabled much, much quicker. And what about if you’re not sure if you need a network element, or not sure how much of the resources of that node you need? NFV is a great way to get started. If you don’t know how many resources you need, then you can get started using this and scale it up relatively easily. 

Is there a clash of technologies between IT and telco when it comes to NFV and SDN? My opinion is that NFV is a step on the way to SDN. They can co-exist within the same infrastructure – for example, the network operator data center that feeds into the enterprise network. They do not compete.