Performance Evaluation of Two Service Recovery Strategies in Cloud-Native Radio Access Networks

Abstract

Through the virtualization of network functions (VNFs) Cloud-native radio access network (C-RAN) architectures bring several advantages to mobile virtual network operators, such as reduced cost, improved flexibility, scalability, and security. However, when these VNFs are affected by Cloud failure or maintenance, the service downtime perceived by the C-RAN mobile user should be minimized whenever possible. This study investigates experimentally the C-RAN mobile user service downtime in the presence of a failing mobility management entity (MME) when two alternative recovery mechanisms are applied: i) a live migration of the virtual MME from the affected data center to a backup data center, and ii) the use of a pool of geo-diverse MMEs, which is a functionality available in S1-flex supported by OpenAirInterface. The study leverages a Software Defined Networking (SDN) optical backhaul network that consists of Open ROADM equipment configurable through the TransportPCE Controller and PROnet resource Orchestrator. The former mechanism should be used as a proactive recovery and therefore activated when the MME failure is predictable, whereas the second mechanism should be used as a reactive recovery and therefore activated to overcome sudden and unexpected MME failure occurrences.