Dreamin: Channel-Aware Inter-Slices Radio Resource Scheduling for Efficient Sla Assurance

Network slicing addresses Quality of Service (QoS) needs through efficient inter-slice Radio Resource Scheduling (RRS) to meet Service Level Agreements (SLAs) of each slice. Given the independence of each slice's scheduler and the discrepancy in channel quality among users and resources in different frequencies, an effective inter-slice RRS must predict intra-slice RRS and be channel-aware. In plentiful scenarios, where all SLAs are assured, RRS can also improve energy efficiency by reducing allocated resources, while in scarce scenarios, where disrespecting SLA is inevitable, we can reduce the SLA drift (SLAD), a metric quantifying SLA non-compliance. In this context, we formulate a constraint programming problem for channel-aware RRS oriented to SLAD and resource usage minimization and propose the Drift and Resource Allocation Minimization (DREAMIN) scheduler as a scalable approximation of the solution. Our simulated results show DREAMIN outperforming the state-of-the-art RadioSaber (RS) reducing the resource usage by 62% in the plentiful scenario and lowering by 62% the SLAD in the scarce scenario. DREAMIN also has 14 times more occurrences of high intra-slice fairness indexes than RS in the scarce scenario, demonstrating how schedulers oriented to minimizing SLAD tend to be fairer than others maximizing total capacity.