Previously published on Light Reading
Arguably there are three key capabilities that operators are looking for from NFV service assurance solutions: 1) virtualized active probes for service testing; 2) virtualized passive probes for service monitoring; and 3) a near-real-time, big data collection and analytics engine. This latter aggregates probe data and analyzes it for troubleshooting and performance management and helps determine corrective actions (self-healing) such as scaling or migrating VNFs. For all this to work, NFV service assurance must be integrated with the service orchestrator responsible for managing the VNF lifecycle. Service assurance analytics is a key input to the orchestrator, driving remedial changes to services/VNFs.
VNFs spun up to create a virtual network can't be pre-tested in a lab and then released to the production network since they only come into existence on-demand. Operators need to be able to automatically test VNFs in real time at the time of instantiation. Since the service model dictates which VNF instances will be spun up and where, operators can use service models to guide the active testing of the service throughout its lifecycle without the need for manual intervention. If a virtual network is created on demand for a customer service, then a virtualized active probe, itself a VNF, can be spun up on the fly, as part of that virtual network.
Unlike a physical probe, which is usually deployed at a network aggregation point, a virtual probe can be placed at the closest location to the end customer, usually the access device. Here it is in the best position to test service SLAs from the customer's point of view. The active probe can provide a service-specific, real-time understanding of the customer's experience. The virtual probe can remain in place for the lifetime of the service, continuing to test and confirm that the service is in order. When the service is torn down, so too is the virtual probe instance.
Virtualized probes are inherently more cost effective to deploy widely in the network than hardware-based physical probes. Deployed on a per-customer basis they can provide high visibility of service performance, enabling operators to control customer experience more effectively, accelerate the identification of problems and take proactive steps before performance is affected.
Ihab Mahna, solution architect at EXFO, sees three stages in the evolution of service assurance to support SDN and NFV: virtualization, automation, and DevOps. In the first phase, virtualization, the focus is on infrastructure assurance. As VNFs are instantiated they are validated during the onboarding phase and then monitored on an ongoing basis. In the second phase, automation, the focus shifts to service activation and monitoring. Service turn-up validation test routines and SLA monitoring are incorporated into the VNF service chain thereby reducing manual test processes. In the third phase, DevOps, real-time assurance insights are incorporated into a DevOps process. Service assurance enriches the DevOps process by embedding simulations, service operations, trouble management, and service validation into the matrix of day-to-day activities. The architecture required to support NFV service assurance is likely to be multi-layered:
The virtualization of active probes is a lynchpin of service assurance for the virtualized network, but it is not the only capability operators will need. Passive probes and an analytics solution that can crunch the data associated with millions of customers and their services are also critically required. Those operators that master NFV service assurance early will be able to reap NFV benefits faster and de-risk VNF-based service launches.