A multi-stage graph based algorithm for survivable Service Function Chain orchestration with backup resource sharing

Network softwarisation introduces flexibility in network management by enabling the deployment of network functions as software modules running on virtual machines. However, this creates new concerns for service availability and reliability due to multiple sources of failures at both software and hardware levels, potentially resulting in service degradations and penalties due to Service Level Agreement (SLA) violations. The survivability of critical services can best be guaranteed by pro-actively provisioning dedicated backup resources for these services, but at the cost of a high resource consumption. Aware of the divergent requirements of future services, a promising alternative is envisaged, allowing non-critical users to use the unused backup resources of high priority users. However, this approach poses a stringent challenge if a critical service disruption occurs, requiring the computation of a traffic rerouting solution for the non-critical requests when preempted from their borrowed resources. In this paper we first propose a generic multi-stage graph based algorithm as an alternative algorithm for Service Function Chain (SFC) deployments. Simulation results demonstrate that the proposed algorithm is optimized in terms of resource utilization, resulting in a 10% improvement in terms of acceptance ratio compared to a given state of the art algorithm, and within a 4% margin of the optimal solution. Based on the mentioned algorithm, we propose a new migration-aware algorithm for the mapping of non-critical services, enabling the non-critical services to borrow the unused backup resources from the critical services while minimizing the probability of preemption they could experience. The migration-aware algorithm results in more than an 8% resource saving, in most scenarios, compared to a dedicated backup strategy, and more than a 70% performance improvement in terms of the number of service preemptions, compared to a cost based algorithm. Additionally, whenever low priority users are preempted from their borrowed resources, we propose a new QoS-aware global-rerouting algorithm for remapping those users, reducing the impact of the service interruption thanks to avoiding the migration of surviving VNFs and virtual links when feasible. The proposed algorithm is shown to outperform a service restoration strategy based on local rerouting, in terms of successful service restoration and resource consumption.