By Kieran Mulqueen, Michael O’Hanlon, Marcin Spoczynski, Brendan Ryan, Thijs Metsch, Leonard Feehan & Ruth Quinn, Intel
Virtualization of Cable Modem Termination Systems (CMTS) provides a large range of benefits for operators within the cable industry. Ensuring that this virtualized CTMS architecture can meet required capacity and performance objectives, both current and forthcoming, provides essential assurance to operators in determining whether to adopt this approach. Typically, human-derived policies are being used for the management of the virtualized CMTS (vCMTS), however these are generally static and can result in over- or under-utilized resources and, eventually, in missing service objectives.
The introduction of Network Functions Virtualization (NFV) paves the way toward autonomous management of Virtualized Network Functions (VNF) like vCMTS. Autonomous NFV management using machine learning (ML) shows huge potential to optimize such a system, benefiting both service providers and customers. Machine learning allows for the generation of models that ensure reliability and dependability of the overall system while minimizing running costs and improving service assurance.
To generate analytic models capable of enabling autonomous management capabilities, the NFV Management and Orchestration (MANO) components need to be presented with key behavioral insights of the VNFs and infrastructure resources.
Here we describe an approach that allows derivation of such insights, which facilitates its use in various models that can be used in a Network Functions Virtualization Infrastructure (NFVI). We also show initial proof points demonstrating areas where machine learning can be used for immediate effect.