Parallel discrete event simulation in star networks with application to telecommunications

We propose a new conservative, time-window protocol for parallel discrete event simulation where the simulation system and the physical system used can both be modeled as a star network, in which one central node is connected to every other node. The success of our algorithm is also dependent of the lookahead predictability of the system to be simulated. The entire approach is designed keeping telecommunication applications in mind. The synchronization algorithm presented in this paper is deadlock free and will not lead to any erroneous computation. We have also computed critical system parameters such as queue length at individual nodes, approximate service times, waiting times and the rate of progress of simulation