Cell loss estimation in an ATM multiplexer loaded with heterogeneous on-off sources

In an ATM network, estimating cell loss probability is a crucial factor in congestion control and traffic management. This thesis proposes a method to estimate the cell loss probability in an ATM multiplexer with heterogeneous traffic sources. We consider an ATM Multiplexer with a finite buffer of size K (cells) and a single link with capacity C (cells/second). Unlike previous work, we approximate the heterogeneous arrival process using a Markov Modulated Deterministic Process (MMDP). We model the ATM multiplexer as MMDP/D/1/K queueing system, then we derive the cell loss probability formula for an ATM multiplexer, and we provide an algorithm to compute it. We consider several classes of traffic sources: voice, low-speed data, high-speed data and image-retrieval. Then, we compute the cell loss probability based on the analysis derived. By simulating the same network, we calculate the cell loss probability. Comparing the analysis and simulation figures proves the accuracy of our results. After examining these results, we observe that a dominant class is present in the heterogeneous ATM network, and it significantly affects the cell loss probability when extra connections are accepted from this class. Moreover, we demonstrate that our model performs consistently well for large buffer sizes where the burst level congestion is dominant. Therefore, this model is suitable for use in preventing burst level congestion from ATM heterogeneous networks