Stochastic traffic features and queueing working conditions: a simulation study

The paper presents an extended simulation study aimed to the evaluation of the impact on queueing performance of different statistical features of actual traffic data. In particular the relevance of marginal distribution (PDF of the arrival rate) and memory properties (LRD-SRD) is considered, highlighting the impact of these features in different queueing scenarios, varying queue length and loading conditions. The study is carried out by means of discrete event simulation driven by actual traces (collected during measurement sessions) and synthetic sequences generated according to different models. The choice of the generation procedures is motivated by model adequacy in capturing the considered statistical features of actual traffic. The analysis emphasises the need of a proper methodology (which takes into account working conditions) in model selection for realistic queueing performance evaluation.