Inter-Slice Radio Resource Allocation: An Online Convex Optimization Approach

Inter-slice radio resource allocation (IS-RRA) is a new layer of radio resource management introduced by network slicing in 5G and beyond mobile communication systems. Instead of focusing on the per-user service quality as in conventional packet schedulers, IS-RRA is required to ensure the service-level agreements on a per-slice basis, which is particularly challenging due to the diverse requirements and behaviors of network slices. In this article, we analyze the inherent limitations of the existing model-based and data-based approaches and propose a novel framework based on online convex optimization (OCO). Specifically, the proposed OCO approach is able to incorporate the offline knowledge and the online data into a general online learning framework, which overcomes the modeling difficulty and high computational complexity of the model-based approach, and avoids the blind exploration of the data-driven approach. At last, we provide the simulation results of an example scenario, which shows that the proposed OCO approach can adapt to diverse service requirements and provide comparable performance to the optimal solutions given in hindsight.