Individual cell loss probabilities and background effects in ATM networks

An asynchronous transfer mode (ATM) multiplexer loaded with two classes of on-off sources is considered. Applying the method of Markov modulated deterministic processes (MMDP), a simple algorithm for computing the individual cell loss probability for each class is developed. With the help of the algorithm, the authors investigate how different traffic streams interact with each other with respect to individual cell loss probabilities. Emphasis is placed on the phenomenon of background effect which may be observed when unbalanced traffic streams are multiplexed into a common output buffer. Its implications to call admission control and bandwidth allocation in ATM networks are discussed