Minimum-Cost Virtual Network Function Resilience

International audience; In the future 5G networks, a wide range of new services with strong requirements will be delivered in the form of chains of service functions on independent virtual networks. These virtual networks will be deployed on demand, each one adapted to the specific service requirements. For infrastructure providers a real challenge consists in providing and setting up the required virtual networks (network slices) while guaranteeing strict Service Level Agreements. One of the major stakes is to be able to provide failure protection for the service function chains at minimal cost. In this work, we consider a set of deployed service chains, and we study the best strategy to protect them at minimal cost. We propose mathematical formulations that provide optimal backup functions placement over a network, and the associated backup paths for each VNF of all the chains. We develop an efficient ILP-based heuristic relying on a separation of the problem into smaller ones to solve large scale instances. We show that our heuristic is competitive, both regarding the solution quality and the solving time.