Hybrid technique for analysis of multiplexed power-law traffic in broadband networks

Research has shown that the simulation of traffic in many broadband network scenarios involves excessive time consumption, owing to the power-law nature of much of this traffic. As an alternative, analysis will often yield systems of equations that are prohibitively complex to solve. Thus, there is growing interest in combinations of simulation and analysis, i.e. hybrid solutions. Here, a hybrid technique is developed for the evaluation of queue state and overflow probabilities when buffering multiple ON/OFF sources, where ON and OFF periods are Pareto distributed. It is shown that this technique allows these probabilities to be predicted accurately, while avoiding insoluble systems of equations and excessively long simulations