There is a significant and growing interest from the service provider community for network functions virtualization (NFV) types of solutions. Service providers see the potential and realized benefits from the cloud and software defined network (SDN) movements and are looking for ways to incorporate the positive attributes of virtualization technologies in their core network solutions. While NFV depends on the virtualization of applications and services, virtualization alone is not enough to leverage the full potential of the NFV architecture.

NFV offers the goal of a simplified and more agile network with the flexibility to adjust to changing conditions. This architecture has the ability to reduce capital expenditures, operational complexity, and improve the total cost of ownership of the network infrastructure for the service provider. As mobile networks continue to transition to an LTE-based all-IP infrastructure, the use of a common architectural framework becomes a realistic goal.

It is important to understand that virtualization alone increases the cost and complexity of the service provider network, but is an essential step to realize the full potential of NFV. The utilization of common off-the-shelf (COTS) hardware and application of software versions of key network functions such as firewall services, load balancing, session border controllers, only delivers a foundation for the service provider to derive monetary and operational efficiencies.

First Steps are Painful Ones

Virtualization alone is not enough to leverage the full value of the NFV architecture. Virtualized network functions (VNFs) residing on COTS hardware are usually more expensive and harder to manage than their vendor-proprietary counterparts.

On the surface, the reduced cost for the COTS hardware delivers a CapEx benefit, but when the cost of the virtual licenses is included along with the performance characteristics of the COTS infrastructure, the performance per dollar cost looks much less appealing. Even though the COTS hardware may cost less than purpose-built hardware, when the additional burden of the virtual infrastructure management is added to the performance capabilities of the COTS hardware, it is hard to design a network architecture where there is a cost/performance benefit by virtualizing the solution.

From an operations perspective, virtualization makes the network more complex to manage with the addition of the virtualization management layer. There is an additional layer of infrastructure that must be understood and managed for every VNF that resides on the NFV infrastructure. While there is a benefit to managing a consistent hardware infrastructure platform, each VNF will still be provided by individual vendors. The VNF infrastructure requirements and their interaction with the virtualization layer may vary, increasing the complexity of the architecture that must be understood and managed.

In addition, while the use of COTS hardware simplifies operations management by offering a common NFV infrastructure, it does not make up for the dynamic nature of the NFV network architecture. The ability to make resources available on demand through the elasticity of the virtual environment means that VNF forwarding graphs will change every time VNFs are enabled and disabled based on network requirements. This becomes an operational concern as it becomes much harder to troubleshoot and understand the network architecture for any given point in time.

Virtualization is Only the Foundation

Virtualization is the essential foundation to extract the benefits of the NFV architecture. Virtualization enables the abstraction of services by separating the function from the physical infrastructure. Technologies leveraged by cloud networks such as on demand computing through elasticity and global server load balancing (GSLB) require virtualized services that can be made available easily and efficiently.

By deploying a COTS infrastructure, VNFs can be made available at any location where the NFV infrastructure has been established. These VNFs can be enabled and disabled depending on resource requirements and based on location, subscriber profile, type of service required, and time of day.

The availability of these VNFs as they are spun up and down within the NFV infrastructure requires the use of load balancing to create pools of resources. These individual load balanced groups are managed through intelligent DNS services via GSLB to steer traffic to the best situated VNF pool based on delivering the optimal service to the data irrespective of physical proximity.

Management and Orchestration Enables the Ecosystem

All of these functions can be coordinated manually from an operational perspective, but elasticity implies automation. For automation to succeed, it is critical for all of the VNF components to have the appropriate APIs for management and orchestration systems. These APIs allow for the collection of information and actions to be sent that reconfigure the network services based on network conditions and operator defined policies.

The management and orchestration infrastructure allows for multiple, disparate VNFs to be coordinated in step as VNF resources are spun up and down as network demands shift over time. The automated synchronization of multiple elements creates a self-sustaining service provider ecosystem of network elements to reliably deliver the essential services to subscribers. It is the responsibility of the management and orchestration systems to deliver the benefits that virtualization enables.

One Step Back, Two Steps Forward

Virtualization delivers the technological features necessary for NFV to succeed, but more work is required to properly realize the potential benefits of this new architecture. It is the abstraction of the VNFs and enablement of the dynamic nature of the NFV framework through management and orchestration that truly allows this technology to succeed.

The forward-looking vision of the NFV architects allows the service providers to see the full benefits of the design even when there may not be realized benefits with the initial steps required to implement the plan. The goal of the scalable, flexible, and reliable network infrastructure with a fully automated management and orchestration system to apply operator policies is within sight.