Spectrally Efficient Guaranteed Rate Scheduling for Heterogeneous QoS Constrained Wireless Networks.

Next generation wireless schedulers will support increasingly heterogeneous users/devices in terms of their traffic characteristics and service requirements. Particularly challenging is the need to deliver low latency traffic with strict deadlines in a spectrally efficient manner. We introduce a class of wireless schedulers, Opportunistic Guaranteed Rate (OGRS) that exploits the temporal variability in users' channel capacity with a view on maintaining delay guarantees. OGRS meets the user's delay constraints by opportunistically allocating the user the equivalent of a fixed service rate, which given a dual leaky bucket constraint on its traffic will ensure the delay requirements are met. We consider offline policies with access to future channel rates, which establishes a bound to the wireless spectral efficiency. We show via extensive simulations that OGRS can be within 10%-40 % of this bound for a range of delays that were considered. These gains translate to more than a two fold enhancement in eMBB users' throughput, when URLLC and eMBB traffic share resources. Finally, we propose a measurement based admission control strategy for latency constrained URLLC users, so that the network can guarantee QoS to all its users - existing as well as newly admitted ones.