Probabilistic analysis of an M/G/1 queue

The well-known queuing model is (M/M/C) : (∞ / FIFO), is widely used in different applications, given its simplicity of implementation. Unfortunately, this model does not always provide good accuracy in terms of service time estimation. In this work, we provide an empirical analysis to identify the probability distribution that is most suitable for modeling the service time of a typical Central Processing Unit. Our experiments indicate that the service time in question is best characterized by an Erlang distribution denoted by Erl7 (0.12).