Abstract
Recently the FCC released 3.85GHz of licensed spectrum and 7GHz of unlicensed spectrum in the frequency band above 24GHz. This massive spectrum can potentially unleash enormous information flows. Transmissions in this millimeter wave band are directional and not subject to typical omni-directional interference, but are subject to absorption.
So motivated, we consider unreliable multi-hop wireless networks carrying many flows, where each flow has an end-to-end deadline. Define the "timely-throughput" vector as the throughput vector of packets that meet their respective end-to-end deadlines.
Several fundamental questions are of interest. What are the maximal timely-throughput vectors that an unreliable multi-hop wireless networks can support? What are the optimal scheduling policies? Since optimal scheduling can require knowledge of instantaneous network state, which in turn requires communication with zero latencies, giving rise to a chicken-and-egg conundrum, is there any hope that fully distributed policies not requiring complete network state can be optimal?
We introduce a price-based scheduling policy for scheduling multiple flows with hard end-to-end deadlines, over a network of unreliable links, with average power constraints at nodes. This policy is easily computed and fully distributed, and it optimizes the throughput vector subject to the hard end-to-end deadlines. We also characterize the set of all maximal timely-throughput vectors.
[Joint work with Rahul Singh]