Ad hoc distributed simulations are constructed by embedding online simulations into a network of sensors. Recent work examining the effectiveness of ad hoc distributed simulations in modeling open queueing networks has been limited to a specific network configuration. This paper examines the accuracy of the ad hoc approach more broadly for open queueing networks. First, we identify conditions where the ad hoc approach will yield accuracy comparable to traditional replicated simulations. For cases where these conditions do not apply, we propose a "buffered-area" mechanism to improve the accuracy of system performance estimates. Empirical evidence is presented showing that this mechanism helps achieve a substantial bias reduction with a moderate increase in execution time.
Published in:
Principles of Advanced and Distributed Simulation (PADS), 2011 IEEE Workshop on
Date of Conference: 14-17 June 2011