In the thesis we are trying to determine how the distribution of request inter arrival times affects the waiting times of requests in the service process. We considered three different distributions of inter arrival times, namely deterministic (D/D/1 service model), exponential (M/D/1 service model) and long-tailed - Pareto distribution (PA/D/1 service model). Thus, we built three models of traffic generators, from which the requests were passed into the service process into the queueing node model, and each model was simulated at different intensities of request arrivals.
We built the models in the OMNeT++ simulation environment, and then we performed simulations and analyzed the results. The results of the simulations are graphically presented with the waiting times for request service throughout the simulation and with a table of average request waiting times for all combinations of models and request arrival intensities.
Since the generation of traffic using the deterministic probability distribution is idealized, we focused on the exponential and Pareto probability distributions. The former is considered to have been most commonly used in the past for modeling telephone and computer networks, while the use of the latter is becoming increasingly prevalent [1-3].
|
