Orchestration for Network Functions Virtualization (NFV) is essential to achieving the goals of agility and service innovation. However, some operators take the view that they can omit orchestration software and use manual orchestration instead, especially for experimentation and/or low-volume deployments. Even in these scenarios, manual orchestration is a bad idea.

What do we mean by Orchestration?

NFV relies on virtual resources such as virtual machines (VMs), storage and networking. NFV also requires an ability to control physical resources in order to implement a complete end-to-end service. Orchestration handles the allocation and management of these virtual resources using software packages such as OpenStack and VMWare, and the control of physical resources using network controllers.

Full-featured carrier class orchestration software provides the basic operations described above along with more advanced functions such as lifecycle management, resiliency and scaling. These features are essential for realizing the benefits of virtualization, especially for dynamic services deployed at scale, like virtual CPE.

Note that the ETSI NFV ISG defines the notion of Management and Orchestration (MANO), which includes both orchestration proper (in the form of an NFV orchestrator) and an adjacent NFV manager. In this discussion I will lump these together as an “orchestrator” for the sake of brevity.

Low volume means no need for orchestration software, right?

Absolutely not.

Even when the volume of virtual elements and/or the rate of change is low, there are very good reasons why orchestration software is needed. Here are some of the most important:

• Preventing Configuration Errors. Manipulating virtual infrastructure is complicated and error-prone. For example, even a small virtual Evolved Packet Core (vEPC) instance can involve many virtual networks, virtual machines, virtual NICs, VM images, VM flavors, security keys, metadata, configuration procedures, connectivity to external networks, floating IP addresses, virtual storage volumes, affinity relationships, alarm thresholds, to name a few. Without the automation of an orchestrator, trying to correctly and consistently configure all of these components is an accident waiting to happen — an accident that will impair critical services. A very common error is the double-booking of resources, which can be prevented by the next item.

• Resource Management and Enforcement. An orchestrator can assure the proper allocation of resources, including no double-booking, enforcing the use of only supported NFV configurations (service templates) and per-tenant quotas.

• Monitoring and Analytics: A full-featured orchestrator is aware of the virtual resources assigned to a service and its VNFs. It can gather information on these resources and apply analytics to enable interesting applications such as the following item.

• Auto-scaling. A full-featured orchestrator can provide closed-loop monitoring and scaling of elastic virtual components. There is really no manual equivalent to auto-scale.

• Remediation of Faults. A full-featured orchestrator provides not only monitoring of virtual infrastructure, but the ability to move functions from failed resources to standby assets. Manual orchestration means manual remediation of faults.

• Integration into OSS/BSS Systems. An orchestrator provides advanced APIs for integration into higher level systems, enabling attractive characteristics such as flow-through provisioning (possibly initiated by a customer at a self-serve portal) enables a multi-tenancy model which abstracts sufficient complexity from NFV so as to allow self-service administration (by end customers) of NFV services.

• Full Lifecycle Management. An orchestrator can provide a single interface (to humans or other machines) which facilitates true end-to-end and full-lifecycle NFV service management

It’s about automation, control and assurance

A full-featured carrier class orchestration solution provides value from the very beginning of the virtualization process. That’s true whether you are looking at trials, low-volume infrastructure or mass deployment. Let’s leave manual orchestration to symphony conductors.