Lean NFV: Light at the End of the Tunnel or Mere Industry Initiative?

The Open Network Summit (ONS), organized in April 2019, revisited what can be considered a long-established but promising technology: Network Function Virtualization (NFV). Lean NFV is a new reference architecture that aims to eradicate the complexity that characterizes the traditional approach launched by the European Telecommunications Standards Group (ETSI) NFV Industry Specification Group (ISG) in 2012. Lean NFV arrives on the eve of a collective industry effort to propel NFV forward using cloud tools and new technologies.

Lean NFV identifies three architectural changes as the underlying pillars that will bring about optimum results: the NFV Manager, the computational infrastructure, and the Virtual Network Functions (VNFs), all of which correspond to NFV Management and Orchestration (MANO)/VNF Manager (VNFM), the NFV Infrastructure (NFVi), and VNFs in ETSI’s terminology. However, Lean NFV introduces two new elements, namely a Key-Value (KV) store and an OpenStack-like plug-in approach. The former serves as a single, unified integration point between VNFs, VNFM, and underlying computational resources. The latter promotes harmonization by means of a plug-in, an approach best epitomized by OpenStack and one where each deployment function is mapped to specific resources with own Application Programming Interface (API ). This contrasts with harmonization by means of an abstraction layer, an overarching API layer that presents underlying resources in a consistent fashion to Northbound functionality like orchestration.

ABI Research analyzes the NFV’s industry status through the lens of Systems Theory, a discipline in which systems always produce in accordance with the goal(s) we institute from the outset. It thus follows that we should be careful in what we ask systems to produce but, pertinent to NFV, the most pressing question is what did ETSI ask its NFV system to produce? And by extension, was ETSI’s intent the most optimal strategy for a large-scale telco upswing?

NFV, as an ETSI undertaking, was established, with the aim of supplanting hardware-based network appliances with software-based equivalents (hosted functions). The initial discussion did not establish whether those functions would be single- or multi-tenant, but ETSI’s initial focus on Virtual Customer Premises Equipment (vCPE) directed NFV momentum toward single-service, single-tenant purposes. Furthermore, another key design decision with NFV was the bottom-up focus that telcos in general (and ETSI in particular) take with new technology implementations. NFV’s narrowness is therefore twofold: one, a function-specific focus that does not offer a clear path to the unified, common platform-driven (telco) cloud; and two, a parochial strategic outlook with no broader view of how low-level implementation(s) can potentially add up to a common platform that can ultimately be the foundation for new commercially viable applications. Though component complexity—particularly the tie up with existing legacy infrastructure—remains an issue, and one that Lean NFV aims to address, the core of the matter remains scope and scale.                            

Lean NFV proposes to kick-start the inception of a new NFV ecosystem driven by technical enhancements, namely a plug-in approach for integration and an open source KV store for harmonious coordination within NFV solutions. ETSI’s goal was to supplant hardware appliances with software-driven, hosted functions for cost reduction. The approach is similar to how system hierarchies evolve, namely from bottom up: from the pieces to the whole, from individual to collective, from actual production to management of production. The aim of a hierarchy, however, is for the upper layers to serve the purposes of the lower system layers. For example, software on the higher layers to emulate the function of customized and specialized hardware. By contrast, ETSI never had an overarching goal that was aligned with (individual) low-level pieces, such as the virtualization of individual elements. On that front, Lean NFV does not move the needle—yet.

A noteworthy technical enhancement in Lean NFV is the KV store, a single point of central control and integration. This constitutes progress in terms of achieving a highly functional NFV ecosystem, not least because it achieves coordination toward a large-system goal (i.e., end-to-end platform) while affording enough autonomy to keep all subsystems (individual VNFs) flourishing, functioning, and self-organizing. This alone, however, is not sufficient for Lean NFV to provide a large-scale commercial upswing. The industry’s wait for that to happen is tantamount to Adam Smith’s “invisible hand” elevating individual decisions to add up to the good of the whole. In other words, initiatives that are characterized by short-term outlooks at their inception, but with results that yield little or no commercial value at the aggregate level. ETSI NFV is certainly one of those projects, but one that can be strengthened if Lean NFV presses the right buttons and avoids suboptimization.

When a subsystem’s goal dominates at the expense of the total system’s goals, the resulting behavior is called suboptimization. Change for the industry on the NFV front will stem from stepping outside the current, parochial NFV commercial undertakings that focus on specific appliances and getting an overview. The commercial world is awash with examples of industry stakeholders acting rationally in today’s best interests but with no prospect of commercial results tomorrow. This phenomenon is attributable to what World Bank economist Herman Daly calls the “invisible foot,” or what Nobel Prize-winning economist Herbert Simon calls “bounded rationality.” From a wider perspective, the industry should coalesce so that current NFV goals and incentives add up to results that center around a long-term optimum, and that, in ABI Research’s view, is almost certainly a telco cloud platform resting on the future 5G core architecture.

Telco cloud is undoubtedly the future for Mobile Service Providers’ (MSPs’) undertakings into new growth ventures. On the one hand, Software Defined Technology (SDN) is bound to play a critical role in connecting data center elements and edge platform to a carrier’s cloud. On the other hand, six years into NFV and NFV’s role in carrier cloud deployment, or if it will play a role at all, has yet to be identified. Analogously, 5G could be a driver for NFV but, given the narrow scope of NFV that targeted customer-specific, site-specific deployments, the role 5G could play is questionable. This is a result of the fact that 5G services and networks are going to be multi-tenant in nature from the onset. Lean NFV could potentially change the fortunes of NFV, but for it to substantially alter the somewhat wrong direction that the original NFV bus took, it is going to need to propose much more than mere technological enhancements such as the KV store.

NFVs original intent was to reduce Capital Expenditure (CAPEX) by substituting custom network appliances with commodity-server hosting of VNFs. That did not materialize, as the operational costs of VNF onboarding and lifecycle management offset the equipment savings. Furthermore, the original scope omitted operations automations, a key strand which would have aided in avoiding sub optimization. To make amends, there is now a collective voice in the industry that expects cloud-native architectures to plow into new commercial territory. Indeed, this is probably the single most important prism through which to examine whether or not NFV will succeed.

In Systems Theory, a powerful way to influence the behavior of a system is through its goal. Similarly, the industry at large must set a broad NFV goal, break it down into specific sub-goals, and work through the details of how to achieve each. Planning for scale and thinking broadly from the start is required to create a commercial opportunity that has capacity and is repeatable. A cloud native vision, and technology in general, is certainly a plausible enabler, but one that would have to have been integrated into NFV from day one. Absent that, all subsequent industry initiatives, including Lean NFV, need to step up by instituting a direction-setter (the goal) that adequately measures progress, reflects a wholesale view of telco operations, and, eventually, either delivers new revenue or achieves tangible savings. NFV is not there yet, but Lean NFV has the chance to shed light by focusing on seeking the right strategic aim, and, this is important, what whole-ecosystem purpose (goal) that aim serves.