Online Convex Optimization for Dynamic RAN Slicing with Quality of Service

With the emergence of 5G and beyond networks, simultaneous resource allocation to a diverse range of services and applications across various industries has been a trending topic. Achieving efficient resource allocation for different services and applications requires a unified abstraction of available resources, which is commonly known as Network Slicing. However, allocating resources among slices becomes a non-trivial problem due to the limited resources and unpredictable requirements of different services. In this paper, we study a dynamic RAN slicing framework incorporating multiple base stations, where workload distribution, radio spectrum and computing resource allocation decisions are made to meet diverse Quality of Service (QoS) requirements. Specifically, delay-tolerant and delay-sensitive services are considered. The unit resource allocation costs are considered to be time-varying. Furthermore, the unit resource allocation costs and QoS requirements are considered to be unknown before making workload distribution and resource allocation decisions. We propose an Online Convex Optimization (OCO) approach for the RAN slicing framework in order to minimize the overall cost and satisfy the QoS constraints in the long run. Simulation results and comparison with two baselines demonstrate that our algorithm outperforms the baselines while satisfying QoS constraints in long-term.