Network Functions Virtualization (NFV) has created a lot of buzz, but it has also created a lot of questions about real deployments and real benefits. I have some good news: NFV is now real, and is starting to roll out. To get there we in the industry had to meet some tough requirements. Here’s how we addressed them.
Beneficial Cost Model
I believe that the real benefits of NFV are related to agility and service innovation, but operators need to see a cost benefit today in order to move forward.
The first cost consideration is how software is licensed. Operators need lower price points, and they also need flexibility to support models such as elastic services and “try before you buy.”
The next aspect is to be able run on right-sized standard servers. That gives operators the ability to pick the best supplier based on cost and execution. It also emphasizes the need for performance on low-end servers.
Assured Performance
Today’s appliances have been optimized for performance in terms of throughput, low latency and jitter, high capacity for calls or sessions, etc. Simply porting the software from a closed appliance to a virtual machine means a big loss in performance.
To achieve the required performance for virtualized applications means that the software needs to be re-designed or re-factored to fit into a cloud environment in terms of elasticity, resilience, platform independence and packet throughput. This usually means reworking the software to use acceleration techniques such as Intel’s DPDK.
True Openness
Some of the early implementations of NFV were software running on standard servers, but in a closed system. This is not consistent with a true cloud model, where much of the benefit comes from its use of software components from an ecosystem of suppliers, all running on open servers and common virtualization infrastructure.
Achieving true openness means that a set of VNFs from collaborating or competing suppliers can be assembled to deliver a virtualized service. It also means that the function of the service is tied to the VNFs, so innovation means new VNFs – not new infrastructure. ACG recently published a report on the business value of such a common platform.
Systems Integration
Automation is required to realize the benefits of NFV, and that means integrating new virtualization systems into existing systems. This process never goes quickly, and the presence of net new software makes the problem more difficult.
The key to easing this process is the availability of open APIs to support zero-touch commissioning of nodes and flow-through provisioning of services. I previously wrote about this topic at SDxCentral.
End Result? NFV Got Real
We at Overture have been working with others in the industry to make NFV real. You can see evidence of this in our announcement with Masergy detailing its innovative new Virtual f(n) service. We have taken NFV from the lab to the field – and we’re not done yet!