On the performance of discrete-time queues with probabilistic server availability

Abstract

This paper considers a problem that arises in the context of statistical, state-dependent resource allocation in communication networks. Specifically, the problem concerns a single-server, discrete-time queue with state-dependent probabilistic server availability/interruptions, and is motivated by the analysis of ground station queueing dynamics in a broadband satellite environment, as well as the analysis of a blocking shared-medium ATM switch with input queueing. The system is analyzed for two forms of probabilistic behavior: exponential and hyperbolic forms. The analysis yields several performance metrics of interest, including average system occupancy, average delay, and steady-state occupancy distribution.