Following on from my introductory post, I wanted to focus this post on NFV.
The one thing that Network Functions Virtualization (NFV) is not short of is points of view. NFV touches many different players. Traditional vendors see their revenue streams being eroded. New entrants see a new market opportunity. Others, like VMware, see a new use case for their core competencies.
For communications service providers (CSPs), NFV is about a transformation that will underpin the future profitability and sustainability of their organizations. The stakes are high!
So when I see architecture proposals that will negate any cost and flexibility benefits of NFV, I get concerned.
The main pitfall comes from focusing transformation around discrete services, leading CSPs to consider the deployment of vertical software stacks as a first step. Figure 1 shows this concept using the ETSI ISG NFV framework as a context.
Figure 1: Software Stacks inhibit NFV Benefits1
I can see the temptation: Aligning implementations to discrete services keeps financial and technical decision-making easy, limits the need for operational change and introduces commercial off-the-shelf (COTS) hardware. A good first step, right? Well…
Let’s think about the implications:
- Hardware can’t be shared between stacks so must be over-provisioned on a per stack/per service basis.
- No cost or agility benefits of shared infrastructures
- The virtualisation infrastucture is locked within the stack
- New and 3rd Party applications are not easily integrated
- Operational processes are largely unchanged
- Each stack is still expensive vendor-specific support services
- It’s a one shot implementation
- Vertical software stacks cannot easily be ‘un-picked’ at a later date to create a common multi-vendor platform
In short, a hardware-based silo has been replaced with a software-based silo with little measurable benefit. Fixing this will require a new and unplanned business case to start again.
Instead, a virtualized service ‘platform blueprint’ should be developed onto which multiple vendor applications can be on-boarded, similar to that in figure 2.
Figure 2: A horizontal, multi vendor virtualization platform gives the CSP control1
This will require cross-service planning, investment and prioritization to first build the platform and then ‘on-board’ the services. This model has huge benefits including:
- Innovation is directly under the control of the CSP
- Service vendors must compete to be selected
- Flexibility increases as new applications and services can be on-boarded directly to the platform
- New services integrate through the platform and under CSP control
- Capital savings grow over time as new applications leverage a single pool of resource
- Multiple applications can leverage a common pool of excess capacity rather than having to maintain that capacity per service
- Capacity planning can be implemented across the entire suite of supported services, rather than per service. As services grow they can be transitioned to services with higher throughput; less demanding services can be moved to lower-cost hardware.
- N+1 redundancy can be applied across multiple services; N+1 x n.
- Support costs reduce as hardware maintenance shifts from 4 hours call out to the routine replacement of blades
To succeed with NFV service providers must embrace, not fight, the principles and opportunities of resource sharing. That’s not an easy transformation to make, but no transformation that’s worth-while is going to be easy.
In my next blog I’ll look at the operational implications of NFV.
- Key to diagrams:
- VNF – Virtual Network Function. Aapplication or applications combining to create a specific network capability of sub-capability. Multiple VNFs would combine to form the services shown.
- NVFI – Network Functions Virtualization Infrastructure. The combined hardware and software that delivers hardware resources as virtual resources to the VNF applications.
- NFV M&O – Network Functions Virtualization Management and Orchestration. A set of management tools that govern the overall delivery of services.
Nigel Stephenson joined VMware in 2014 to champion VMware’s virtualization solutions for carrier NFV deployments across EMEA. Nigel brings to his current role a depth of telecommunications experience built through the the design, delivery and marketing of networking solutions in a career spanning almost 30 years.
You can learn more about NFV and the future of virtualisation at VMworld Europe, taking place in Barcelona October 14-16 – Register Here