When building your D-NFV infrastructure – scale up or down?

As the network functions virtualization (NFV) world is currently expanding from the data center into branch offices, service providers are looking at offering managed services using virtual customer premises equipment (vCPE) hardware platforms.

For service providers which have implemented NFVs in the data center and chosen an Openstack-based VIM, the road into SMB (small and medium business) premises looks simple: install the Openstack in a compute node at the customer’s premises.

But is that as easy as it sounds?

The data center virtualization uses complex dual-processor compute nodes at a cost of some $10,000 each.

Is it affordable to install a $10K node in a 20-employee branch that only needs access to the customer VPN and to the internet?

The economics of branch virtualization must be cheaper in order of magnitude.

We should be asking ourselves: can our data center VIM work on the type of hardware that a customer can afford to install in a small office branch?

As software and platform providers for NFV deployment, we at Telco Systems ask ourselves: does it make sense to take a ready-made Openstack VIM and try to squeeze it into the x86 hardware platform that fits the customer needs, or should one be built from the ground up?

Lessons learned from the automotive industry

The software industry has learned a thing or two from the automotive industry. One lesson was “Agile/Scrum,” the development method of choice for the 21st century that we “borrowed” from Toyota TPS (Toyota Production System).

Another lesson is: choose the right platform.

Here is a typical platform selection problem for a carmaker: I have to build a new car, which of my existing platforms can I utilize, or should I create a new one?

Automaker “Acme” has identified a new market for small city cars.

Its smallest platform is the one used to build family sedans and station wagons and it is definitely too big; the car chassis is 177in. in length (4.5m).

Can “Acme” build a new chassis and reuse the rest of the platform?

Apparently not – the engine is too big, the transmission is too big, and the fuel tank is huge. This is a bit like trying to squeeze an elephant into a small car.

This is the wrong platform and the automaker had better design a new platform for small city cars.

So what should we aim for?

This is our target in the automotive industry: a car that is no more than 140in. long (3.6m).

How does this apply to the NFV world?

Let’s remember that we come from a world of data center, which includes entire rooms of hardware communication equipment.

Data Center

In the world of NFV deployment, all this can be inserted into one single box.

This box is only a little larger than your home Wi-Fi router, with an x86 processor as small as 4 Intel Rangeley cores, 16GB RAM and A/C powered.


Now, let’s build the right NFVi (NFV infrastructure) for this platform, from the ground up.

Additional resources

To learn more about how to create the right NFVi for your company, contact us

To set up a demo on our NFVi at MPLS+SDN+NFV World Paris 2017, click here


Related posts