Incorporating truncated exponential distributions in queueing models with adjustable service-rate control

Queueing models with service-rate control provide more realistic simulation results compared to simple M/M/1 models which have too much variability in the queue length. Because the variability in number in queue is increased by the unbounded nature of the exponential distribution, another approach modelers sometimes use is to select a bounded distribution, or to limit the maximum sample value by truncating the exponential distribution. This paper compares the beneficial effect of service-rate control and exponential distribution truncation. Simulation results demonstrate that models incorporating both mechanisms (truncated exponentials and service-rate control) generate the most realistic simulation results when rate adjustment is used to reduce queue-length peaks back to normal.