Dynamic Parameter Substitution for solution of Layered Queueing Networks with Timeout Decisions

Solving performance models using queueing networks poses a challenge because parameters such as service times and routing are fixed and must be known prior to solution. Models which involve decisions based on performance quantities must therefore often be solved using a state-based model, through simulation, or a hybrid combination of these approaches. However, state-based approaches suffer from state space explosion for even moderately sized models, while simulation can be time consuming. This paper broadens Layered Queueing Networks (LQN) to handle a subset of models with state-based behaviour, namely systems with timeouts and aborts. The approach, called dynamic parameter substitution (DPS), updates the parameters of the underlying queueing networks of a LQN model as the model is being solved. This approach in this paper is both fast and highly scalable, compared to simulations (LQSIM and CSIM) and a hybrid solution.