NFV progress demonstrated in the form of market-ready improved Virtual CPEs

With the insane amount of hype around developing technologies like SDN and NFV, one can’t help but wonder if it’s on a fast-track to becoming the most expensive vaporware that has ever graced the tech industry, thankfully service rpoviders are here to ease those concerns with fully functional virtual CPE solutions. Due to the immediate advantages of virtual solutions in large deployments, the first ones to get the technology will always be enterprise customers, but it won’t take long for consumers to start seeing these appliances in their homes as well.

Currently there are 2 opposing architectures ISPs can choose to implement in their virtual CPEs.  There’s the service provider hosted centralized vCPE, and there’s localized vCPE. Both of these different implementations have their pros/cons as well viable uses cases in different segments of the market. Due to the fact that NFV is still a relatively new technology, it is impractical to become tied down by the current limitations in either of these architectures, and that is why a hybrid model that can leverage advantages from both sides has come to be the prevailing architecture. 

All of these different implementations have their pros/cons, below I will explain each type, their weaknesses/strengths, use cases and ideal hardware platforms for proper deployment. 

Definition and use case:

Centralized virtual CPE

This solution is implemented by hosting almost all of the VNF’s (Virtualized network functions) on the service provider’s network. In this instance the only additional hardware purchase necessary is a low-power network appliance that acts as little more than a bridge to the cloud hosted network functions.

Use cases: currently only SMB’s are practical for deployment, as various network functions when scaled up do not perform well in a cloud hosted model, due to issues like latency and network tromboning.

Localized Virtual CPE

This implementation requires potent hardware to run all of the virtualized network functions on-premise.  All the NFV’s are integrated into industry standard network servers that are usually in the form of x86 rackmount network appliances.

Use cases: this can be deployed onto any branch, no matter the size. The only restricting factor is the cost of the network equipment necessary to power it all, which was one of the deciding factors in the switch to virtual CPEs(exceedingly so in the case of enterprise networks in the larger businesses).

Hybrid Model Virtual CPE

This architecture takes a flexible approach and runs many network functions on-premise as well as in the service providers cloud network. This method brings about the widest range of support for network functions and is the preferred architecture in many settings.

Use cases: This flexible architecture was made to ease deployment onto any size company branch. The ability to run essential functions locally and offload other network functions onto the service provider’s network makes it suitable for the ever-expanding networks we see today.

Choosing the right hardware for the different implementations

Now that virtualization has disaggregated hardware and software, one can really appreciate the amount of options software based solutions provide. We now have a much larger list of vendors from which to source the hardware for our Virtual CPE's. When deciding what hardware to deploy on, it’s now become the infinitely easier task of simply looking at all the hardware vendors, comparing them and deciding based on the benefits they provide. Gone are the days of vendor lock-in and incompatibility. You can just as easily choose to deploy the vCPE on capable existing servers with little to no modification.

While one of the biggest merits to virtualizing consumer premise equipment is the ability to run as a virtual appliance, there are still many benefits afforded from utilizing dedicated, purpose-built hardware for vCPE’s. Out-of-band management, QoS, 4g connectivity, power supply redundancy and reliability just to name a few of the top of my head. Footprint and form-factor are also playing an increasingly important role in hardware decisions, not surprising given the fact that real-estate and HVAC aren’t exactly cheap resources.

There are 2 popular form factors for x86 networking hardware, the first one being the well-known x86 rackmount network appliances, and the other are the small form factor desktop network appliances. Both have different roles to play in their respective implementation and provide benefits in several areas that virtual appliances cannot.

Rackmount network appliance:

.

Lanner's FW-8759 - x86 rackmount network appliance for network security, UTM, SD-WAN, vCPE...

With the freedom to source our hardware from vendors of our liking, new whitebox network appliances are starting to come out, well to be frank whitebox is really just a fancy new word for the well-known x86 rackmount network appliances that have been available for over a decade. Having no specific software preference (aside from network tailored functionality), they are a fantastic choice in new deployments. The high throughput capacity, redundant power supply and stackable form factor make for a modern, clean looking CPE. These platforms offer high computational density needed in localized implementations, and even medium and large branch Hybrid CPE deployments.

Desktop Network Appliance:

One of the advantages service providers have with centralized CPEs (and to a lesser extent Hybrid CPE deployments) are the lowered requirements for customer-side hardware. This means it will be easy to deploy their vCPE’s onto low-power, small footprint network appliances like the ones below.

 NCA-1210 small form-factor x86 desktop network appliance 

Don’t let appearances fool you, these small, fanless network appliances contain a capable dual core/quad core x86 processor designed with network applications in mind. By removing unnecessary components like the graphics processor and tailoring it towards networking applications, these appliances are able to strike a great balance between performance, size, price and reliability that make them extremely appealing for service provider’s emerging CPE deployments.  

Units like these will easily be able to run on-site network functions like firewalls and WAN acceleration for hybrid CPE implementation in small branches. They even include options like mini pci modules for 4g wireless cards for added resilience and remote management functionality- even when the network is down.

In conclusion:

In the end it all boils down to choosing the right tools for the job, and thanks to the NFV initiative we are now in a better position as consumers than we ever were. We have been given the freedom to decide what we really want for our networks and that’s not a bad thing, in fact it’s a great step in preparing our networks for the future.