NFV Is Not Just For Service Providers

What comes to mind when you hear the term Network Functions Virtualization(NFV)?  Until recently, at least in my mind, the answers were ‘Service Providers’, ‘vCPE’, ‘Cloud’. If you don’t know what NFV is, read this blog post first and come back.

But why does NFV have to be exclusively for service providers?  At the end of the day, the vCPE resides in the enterprise and so the enterprise is the consumer of NFV.  Do we have to sit and wait for service providers to give it to us?

Why NFV?

Before we go into NFV for enterprises, let’s look at why NFV in the first place.  Why are large service providers looking at NFV? For SPs, NFV is estimated to bring 47% cost savings per site per year and to offer an ROI of 156%^[1].  It will increase their revenue by 1.4 billion USD by upselling and cross-selling existing and new SDN-based services.

NFV promises to bring agility and flexibility to SPs when deploying services to vCPEs for their enterprise customers. And by moving to commercial-off-the-shelf (COTS) servers, it will further reduce hardware costs.  Lower costs and increased revenue.  What’s not to like?

NFV for Enterprises

Enterprises can derive the same benefits from NFV.  A few benefits according to ‘The 2016 Guide to SDN and NFV’^[2] include:

• Lower CAPEX and OPEX
• Rapid service provisioning
• Consistent policies across the enterprise
• Reduced risk through service distribution

Your peers seem to be already going this direction.  According to a survey in the same paper, 52% of professionals believe that NFV has either significant or very significant relevance to enterprise IT architectures.  In this survey^[3], 61% say they are in various stages of researching, testing or deploying NFV.  The reasons they cited as benefits of NFV were #1 agility #2 lower server cost #3 increase capacity through software.

NFV for SPs are still in the early stages.  Transforming their massive network will take years, if not decades.  In the meanwhile, enterprises can and should start your journey to network virtualization.

Here are some considerations:

Smooth transition

Transitioning to NFV will not happen overnight. Overhauling your entire network architecture to NFV will simply not be an option.  You will probably start out with a few locations that will benefit the most from NFV.  Locations that require agility, that may need constant shift of network services, or you could start with locations that are less business-critical.

In any case, your NFV solution will have to support this hybrid model.  Some locations with the traditional architecture and some locations with NFV.  Not all locations will look at same.  You do not want to be running two separate networks.  The networks need to work together, smoothly.

SDN and NFV

In order for NFV to work in an enterprise environment, it needs to go hand in hand with SDN. If NFV is about agility and flexibility, making manual configuration changes with CLI would be counterintuitive.  The orchestration and management of services, devices and the infrastructure should be simple and easy. It should also support the hybrid network infrastructure mentioned above.  You should be able to define policies based on business needs, security requirements and application performance.

SDN has finally taken off in the Enterprise.  Controllers with applications extracts complicated network configurations and uses policy-based approach.  It automates provisioning of the end-to-end network infrastructure.  Because of SDN, NFV can now be deployed in the enterprise and is more consumable, together.

Support system

NFV has many components.  I think, for the enterprises, the problem would be troubleshooting these independent components.  NFV, because of its independence of hardware and software, is a composite of multiple layers.  And it is up to you to determine what and where you get each layer from.  The beauty of NFV is that you can get the hardware from one vendor and software from multiple other vendors. But when you need support on an issue, who do you contact?  Who will be responsible for making sure your solution works seamlessly together?

You might already have deployed some virtual network services (i.e. a virtual WAN optimization solution (vWAAS) and virtual firewall (vASA)).  With Cisco, these can be deployed on the UCS E-Series inside the ISR.  But on a hypervisor that was built for applications and not network services, these solutions are not holistic.  It’s not optimized for network services.

And all the layers do not have to be from one vendor, at the end of the day, you may be one of those enterprises who are trying to get away from that.  But you do want a solution that will make it easy for all the different layers to work well together.

So, are you ready for NFV?  I hope this post at least spark the notion.

[1] ‘vCPE services business case: potentially billions of dollars payback for fixed CSPs’ October 2015 Glen Ragoonanan and Gorkem Yigi, Analysys Mason

[2] The 2016 Guide to SDN and NFV by Jim Metzler produced by Webtorials

[3] 2015 Networking PI Worldwide Data by Zeus Kerravala at ZK Research