Project
Development of a methodology to increase the applicability of queueing models whereby numerical calculation of boundary probabilities is required
Queueing theory provides fundamental tools to study the performance of computer
and telecommunication networks. Basically, a model is developed, whereupon
performance measures are deduced by which the scenario at hand can be
evaluated. In a queueing model, broadly speaking, customers (packets, processes)
arrive and are stored in a buffer in awaitance of their service (transmission, being
executed). A popular and effective technique to obtain performance measures,
such as the average number of stored packets, makes use of so-called probability
generating functions. However, in a broad range of scenarios, the resulting formulas
for the performance measures contain various boundary probabilities that have
to be calculated numerically, which is a drawback for applicability. The purpose
of this project is to obtain accurate closed-form (i.e., no numerical calculations
required) approximations for the performance measures in such cases.