Service providers all over the world are in the middle of a shift from hardware-centric to software-controlled, service-focused networks. They're chasing substantial and far-reaching benefits such as reduced costs, speedier service rollouts and enhanced flexibility to address emerging needs. Network functions virtualization (NFV) is one of the break-through technologies enabling the transition.

NFV enables a dramatic change. Network functionality is virtualized, enabling standard servers and general-purpose storage to operate network appliances, which take the place of dedicated hardware. Consider the cumbersome range of hardware types, functions and variants that are currently required over the lifespan of IT services with a given customer, and the substantial cost, scalability and flexibility benefits are readily apparent. Virtualizing hardware-specific functionality eliminates the need for diverse hardware and allows for an unprecedented degree of service agility. Service providers are (finally) freed to profitably offer value-added services to enterprises, enabling those organizations to focus instead on their own core-business concerns.

While many operators initially consider driving NFV from the cloud in order to achieve economies of scale, most are also aware of the need to optimize scalability and performance and to keep latency to a minimum. Hence, a virtualization capability is required both within the network/cloud and at the network edge. Balancing the mix of centralized and decentralized NFV functionality in this way better optimizes network, server and storage resource utilization end to end. It also helps avert lags in performance and the introduction of latency.

In order to fully realize those benefits, however, service providers must take into consideration a host of functional requirements. The shift to service-centric networks needs careful planning with regard to connectivity infrastructure and respective network topologies.

Let's look at the change specifically as it pertains to one key application: virtualization of customer premises equipment (vCPE).

While network interface devices (NIDs) have been deployed for service demarcation in Carrier Ethernet services for many years, a more sophisticated NID is required in the migration to NFV:

• NFV provides the opportunity to run virtual router appliances from central servers instead of installing physical devices on the customer premises. Programmability allows a NID to perform IP forwarding. That means local IP traffic doesn't need to be routed through a virtualized router hosted on a central server.
• As operators extend their business from Layer 2 connectivity services toward value-added managed services, such as offering network appliances, the demarcation technology in the NID needs to monitor higher network layers.
• NID technology has to respond to the extended attack surface of open interfaces and hypervisor-based network architectures by adding an extensive set of security functions.
• Servers integrated with a NID can host multiple virtual machines, which themselves host virtual network functions (VNFs). These can be chained together (just as hardware is physically connected today) for a rich service offering. Because this connectivity among VNFs needs to be established in an automated fashion, NIDs need open programmability.

Migrating from a hardware-based to a service-centric model has profound implications on the functions that must be carried out across the transport infrastructure. Essential functional requirements must be satisfied in terms of balancing centralized/decentralized virtualization, as well as ensuring programmability, end-to-end security across provider and customer networks, and operations, administration and maintenance.

With these transport considerations properly accounted for, service providers are primed to make the long-awaited shift. Through the implementation of NFV they can finally create agile, service-centric infrastructures.