Optimal QoS-Aware Allocation of Virtual Network Resources to Mixed Mobile-Optical Network Slices

Slicing allows 5G networks to accommodate services with different needs over a unified physical network infrastructure. Particularly, in radio access networks (RANs), the baseband processing functions of different slices are treated as virtual applications running on shared compute nodes. Managing this cloud-native network in a cost-effective way requires tightly cou-pled control of connectivity and processing resources. This paper proposes an optimization problem to minimize the overall cost while guaranteeing distinct latency and reliability requirements of different network slices deployed over the infrastructure. This is achieved by choosing for each service, the functional split and transmission parameters that best adapt to the connectivity and processing resources available in the infrastructure. Results demonstrate that the proposed flexible resource allocation scheme provides considerable cost saving (up to 2.4 times) and reduced request blocking (up to 2 times) compared to conventional fixed deployment techniques. The reliability requirements can be guaranteed by packet duplication and/or virtualized forward error correction at the expense of consuming connectivity and/or processing resources, respectively. The flexible assurance of the reliability requirements in the proposed scheme contributes considerably to the achieved improvements on cost saving and request blocking.