Playout buffer responsive wireless streaming for multiple clients

We consider a problem faced by wireless base stations in which multiple requests to stream data must be accomodated while minimizing the amount of buffering time spent by clients. In our model, clients request content in discrete data chunks, but the base station is restricted in which clients can simultaneously be served data at certain rates. This limitation occurs frequently due to wireless connectivity and congestion issues in which some clients are difficult to reach from the base station, whereas others are more easily serviced. These constraints are represented by a set of admissible service rate vectors from which a scheduler at the base station must choose. We take a queuing theoretic approach to this decision problem and employ a stress alignment approach to ensure that maximal throughput from the data center to the clients is achieved. As a consequence, whenever it is possible to stabilize the backlog of requests from each client, we are able to do so. Numerical experiments show that among the variations of this scheduling algorithm, we can choose parameters to vary the priority levels of different clients and even induce dependencies between service rates of different clients.