Probabilistic, Real-Time Scheduling of Distributable Threads Under Dependencies in Ad Hoc Networks

We consider scheduling distributable threads that are subject to dependencies (e.g., due to mutual exclusion constraints) in ad hoc networks, in the presence of node and link failures, message losses, and dynamic node joins and departures. We present a gossip-based distributed scheduling algorithm, called Real-Time Gossip algorithm for Dependencies (or RTG-D). RTG-D builds upon our prior algorithm called RTG, that schedules distributable threads without dependencies in ad hoc networks. We prove that thread blocking times under RTG-D are probabilistically bounded, thereby probabilistically bounding thread time constraint satisfactions’ under the algorithm. Further, we prove that RTG-D probabilistically bounds deadlock detection and notification times, and failure-exception notification times for aborting partially executed sections of failed threads. Our simulation results validate RTG-D’s effectiveness.