Network functions virtualization (NFV) is changing everything. NFV makes the network more flexible, more agile, more programmable. That’s essential, because a service provider can offer the high service variety and quality that customers want, quickly and cost-effectively, with NFV – but only if the end-to-end service is properly orchestrated.

A major challenge is that NFV makes networks and their functions less deterministic. Traditional network services are relatively static in their configurations – measurable and consistent. Load balancers, firewalls, content filters, caching appliances… network operators have an excellent understanding of exactly what those devices do, what types of load they can manage, and how they perform under various load levels.

By contrast, the performance and capacity of NFV-defined functions are transformed by the multi-tenant hosting infrastructure (think network server) and the current load from all its applications. The workload on that infrastructure is constantly changing due to the context of the network at any point in time. Yes, NFV can enable flexibility and agility, but it’s hard to monitor exactly what’s going on and thereby proactively manage it.

That lack of determinism frustrates traditional means of administering the network, because the need for real-time operations support systems (OSSs) is moving deeper into the network itself. In an NFV world, network operators have to dynamically manage the life of each virtualized network function (VNF), in addition to managing end-to-end services. With those VNFs running on distributed hardware with disparate loads, there are three separate challenges:

• Managing each VNF
• Managing the segment of the customer’s service running on that VNF, or VNF service chain, referred to as a “network service”
• Managing the customer’s end-to-end service across the entire network

(Note: There are different definitions of “service” floating around in the industry. For clarity, I refer to a service as complete end-to-end data connectivity for the consumer or enterprise, spanning both physical elements and VNFs.)

The best way to manage these complex NFV-driven networks is with “orchestrators of orchestrators” – a multi-tiered system focused on provisioning and assuring each of those functions.

At the lower level, a Domain Orchestrator (DO) manages the operational requirements of the individual VNF. Specifically, the DO manages the compute, storage, and networking requirements of the VNF. At the higher level, there’s the Lifecycle Service Orchestrator (LSO), which is in charge of specifying, provisioning, managing, and tearing down the customer’s end-to-end services across heterogeneous networks.

I propose that this functionality include what is currently being defined within the ETSI NFV MANO reference architecture as a Service Orchestrator (SO). The SO manages customer-oriented services across various virtualization domains, by marshaling individual VNFs together into a network service. Think of the SO as the means of instructing multiple DO instances how to configure and connect their VNFs across the network.

The LSO not only manages VNFs by controlling, or even including the functionality of, SOs, but also seamlessly orchestrates services across both VNFs and traditional physical network functions (PNFs) that aren’t implemented via NFV. The functions of the LSO are being defined by the MEF, which is collaborating closely with ETSI.

The above might imply that network operators need three separate pieces of orchestration software, requiring complex integration, in order to set up VNFs (DO), set up customer services on those VNFs (SO), and combine them into end-to-end services (LSO).

Fortunately, that’s not the case, as modern-day comprehensive software solutions can implement required orchestration functions using open, flexible APIs, enabling management from a single pane of glass. LSO software can eliminate complexity within heterogeneous networks comprising both traditional PNFs and next-generation VNFs, turning what could be disjointed, time-consuming operations into a streamlined and user-intuitive management implementation.

The upside? Service operators can offer customers flexible, programmable, agile, and scalable solutions, leveraging the capabilities of NFV while taking into account the nondeterministic nature of VNFs within a real-world network. Operators can also leverage the capex savings NFV promises, without being bogged down by opex increases. Implementing orchestrators of orchestrators is a win-win for operators and customers.