Inspired by classical results in real-time scheduling, we consider a model of an access point serving several flows, and provide a possibly surprising characterization of what throughput vectors feasible, as well as throughput-optimal scheduling policies that are almost static. Then we turn to some interesting recent results in multi-hop networks with end-to-end deadlines, and show that under some models of concurrency constraints, one can obtain decentralized policies that are throughput optimal [joint work with Rahul Singh, I-Hong Hou and Vivek Borkar].
