Microsoft has developed a radio resource scheduler that fulfills throughput and latency SLAs for individual apps operating over a cellular network.
Microsoft's system organizes applications with similar SLA requests into network slices. It utilizes traditional schedulers that optimize base station throughput by creating resource schedules for each slice to meet the requirements of each application. In this model, applications convey their network needs to the operator by specifying minimum throughput and maximum latency. Our system, acting on behalf of the operator, ensures these SLAs are met across the shared wireless medium by calculating and assigning the necessary PRBs (Physical Resource Blocks) to each slice.
The systems is able to decouple the network model and the control policy by formulating SLA-compliant bandwidth allocation as a model predictive control (MPC) problem.
Microsoft says MPC is great at solving sequential decision-making problems over a moving look-ahead horizon. It decouples a controller, which solves a classical optimization problem, from a predictor, which explicitly models uncertainty in the environment.
Here is a Microsoft blog:
