Some of the most relevant papers tackling bulk service queueing systems with random service capacity were briefly described, through the analysis of the results given it was found that other authors solutions are difficult to apply. The solution method presented here does not require to fit a theoretical probability distribution for service time, but there is a need to find a discrete distribution matching at least the first two moments of the empirical service time distribution. Recursive numerical convolution of probability vectors must be carried out on a computer program, which is easy to implement. Numerical results are accurate enough for practical purposes.
An important advantage this model has among other approximation queueing model, such as those analysing the system at phase epochs, is that the system is observed at service epochs, having in consequence a substantial saving in computer running time.