Calculation of loss probability in a partitioned buffer with self-similar input traffic [DiffServ network]

In the differentiated services model, provisioning quantitative assured services is a challenging topic, as it requires loss probability calculation for a partitioned buffer. In this paper, we study such a loss analysis problem with self-similar input traffic, which has never been studied in the open literature. The input is modeled as a fractional Brownian motion process including J classes of traffic. Each class has its unique requirement on packet loss probability. A first-in-first-out buffer partitioned with J-1 thresholds is used to provide J loss priorities. Heuristic expressions of the loss probabilities for all the J classes are derived, and simulation results demonstrate that the heuristic expressions provide an accurate estimate for all the loss probabilities over the entire buffer range.