Using discrete event simulation to fit probability distributions for autocorrelated service times

This is the latest version of this item.

Official URL: http://dx.doi.org/10.1080/03155986.2019.1624474

In the literature, it has been shown that autocorrelation in service times has a dramatic impact on performance measures in delay systems. Although some studies deal with systems having autocorrelated service times, a general renewal approximation methodology that can incorporate this information in analytical models has not been previously proposed. In this paper, we develop a discrete event simulation based analytical method to fit approximating distributions to capture the characteristics of autocorrelated service times. Our observations indicate that a single approximating service time distribution fails to accurately predict the behaviour of the delay system. Therefore, we include the server utilization information in the developed method, which, in turn, provides a family of approximating service times. Testing our approximation both in predicting the mean delay and the system size distribution of the original systems with autocorrelated service times proves that it is highly accurate.