Interpretation of Dispatching Policies on Queue Behavior via Simulation and Time Series Analysis

Abstract The behavior of uniformly sampled queues simulated under different dispatching policies are modeled and interpreted using time series techniques. Queue behavior is shown to be adequately described by a first order autoregressive AR(1) model if the job selection discipline does not depend on operation processing time. In those cases where processing time was used in job selection, a second order autoregressive AR(2) model is adequate to characterize the queues. Queue behavior under different selection disciplines is then interpreted via the models as a convolution of the specific selection policy on the basic first come, first served (FCFS) queue. This analysis also demonstrates the inability of linear differential equations to describe queue fluctuation in systems where processing time based selection rules are employed.