Optimising microservice-based reliable NFV management & orchestration architectures

A highly reliable set up of Management and Orchestration (MANO) functionality is crucial for telecom operators in order to support demanding NFV-powered telecom services such as elastic emergency or security services, or other applications requiring fast, dynamic and reliable (re-) provisioning processes. NFV MANO functionality controls the entire life cycle of such services, from instantiation and configuration to monitoring, migrating, scaling and terminating them. In this paper, we introduce tunable and scalable mechanisms that provide MANO with high availability and fault recovery, two reliability facets that are barely covered in the data plane dominated state-of-the-art. The mechanisms use state sharing and distributed load balancing in the context of both a centralised and distributed microservice-based architecture. The proposed mechanisms are unique as they are able to quantitatively characterise the trade-offs between both the degree of reliability and the associated cost in terms of bandwidth and computing power. This feature allows us to introduce a cost function to determine which configuration of the mechanism is optimal with respect to an operator's needs.