Letter: Performance evaluation of analytical models for effective bandwidth allocation in ATM networks

Abstract

AbstractSome promising methods for the analysis of ATM systems calculate the buffer overflow instead of ATM cell loss probability. This is done under the presumption that buffer overflow probability is an adequate estimate for the ATM cell loss probability. These two performance parameters may differ considerably in some cases. For instance, if the mean load offered to a link slightly exceeds the link capacity, then the buffer overflow probability equals to one. On the contrary the cell loss probability may be much less than one. The paper shows how to use these methods so as to compute cell loss and examines in detail their relative performance and suitability for the description of an ATM multiplexing environment.