DCD FOCUS: What is Network Function Virtualization (NFV)?
Sanjeev Mervana: When NFV was introduced in 2012 by the ETSI Industry Standards Working Group, the objective of the group was to develop recommendations on how to define industry approaches to the virtualization of network functions; ie, separating software from hardware and creating ‘virtual appliances’ on a general-purpose network. The network would consist of storage platforms using virtualized data center and cloud management techniques. The virtualized network functions would be located in those data centers, on or beside network nodes and on end-user premises.

There are a variety of benefits delivered via the implementation of NFV networks:
Reduced equipment costs and power consumption through the consolidation of equipment and economies of scale.

Reduced time-to-market through shortening of the typical network operator cycles of innovation.

Availability of using network appliance multi-version and multi-tenancy, which allows network operators to share resources across services and different customer bases.

Targeted service introduction based on geography or customer sets is possible, so services can be scaled up or down rapidly.

It enables a wide variety of eco-systems and encourages openness. NFV opens the virtual appliance market to pure software entrants, small players and academia, encouraging more innovation to bring new services and new revenue streams at much lower risk.

How does it work?
A service provider network infrastructure consists of a widely diverse set of components, ranging from elements that support the actual forwarding of packets, to elements that support the management and operations of the network. Today, most of the network functions execute on dedicated hardware solutions. As a consequence, the network consists of many discrete hardware components, each executing a specific function or group of functions.

Ultimately, NFV transitions some of the network components away from custom hardware and software solutions onto industry standard, virtualized compute hardware. As service providers start deploying virtualized network functions in their networks, the effectiveness depends on the individual use case. As an example, for use cases that need innovative control and signalling functions with low bandwidth requirements, NFV will bring them the required flexibility and quick deployment. However, when highest performance is at the essence for bandwidth-demanding applications based on stable and well-defined algorithms, those functions run most efficiently on application-specific ASICs.

Who is developing it?
Many vendors are looking at the concept of NFV by virtualizing point functions, only addressing individual elements of the network. For Cisco, it’s not about having a point product in only one of these network areas, it’s about creating horizontal solutions and stitching them together.

How does it tie in with SDNs?
NFV is highly complementary to the software-defined network (SDN), but it is not dependent on it, or vice-versa. SDNs enable administrators to control their network away from their hardware through software applications. NFV can be implemented without the requirement for SDN; however, the two concepts and solutions create considerably more value when they are deployed in combination to deliver the customer objective around service velocity, business agility and operational simplicity.

Why do we need it now?
During the last decade, a variety of trends have created an imbalance in the cost for service providers servicing consumer and enterprise customers compared with the revenue generated from those service offerings. Service providers need to lower TCO and increase monetization of new and competitive services simultaneously. The focus is on three areas to achieve this: service velocity, business agility and operational simplicity. By using a framework that combines the SDN, NFV and cloud services, service providers can deliver products quicker and more efficiently than ever before. Embracing this transformation enables operators to scale their networks, launch new services and adapt to changing traffic and business models.

What are the key barriers to adoption?
The reluctance to use NFV in the past has been influenced, in part, by factors such as lower packet processing performance on general-purpose central processing units, unpredictable and low performance of software executing under a hypervisor, close linkages between software and hardware solutions and a comparatively simple operational and support model. All of these factors impacted the key requirements from service providers as it related to network resiliency and scale.

It is important to make NFV deployment and operation as simple as possible. Operations must evolve from today’s model, which supports physical network entities, to a model supporting virtual network entities. These changes will require new functionality in the network management solutions and changes in the operational procedures. NFV should not be viewed in isolation; instead, there are several concurrent technologies, including SDN and cloud, that need to be addressed in parallel to make it a successful evolution.

This article first appeared in FOCUS 31. You can view the digital edition of the magazine here or download a copy for the iPad at DCDFocus.