The cell loss process in ATM networks and its impact on quality of service

Real time cell loss measurements have been carried out for a conceivable ATM network scenario. By measurement, and compared to theory, adequate performance levels were readily obtained in terms of e.g. long-term CLR (Cell Loss Ratio). The long-term CLR does not however provide enough information about the temporal distribution of the cell loss process. In view of this fact we conclude that short-term CLR for a cell loss burst better signifies the deterioration of a service. In the measurements two different types of ATM cross connects were employed. The ATM cross connect with the larger buffer size did not only result in lower long term CLR but also significantly lower short-term CLR. In these measurements the short-term CLR was 3-5% for the larger buffer size (500 cells) and 7-9% for the smaller buffer size (100 cells). The average short term CLR was not significantly influenced by the network load level. The acceptable cell loss level is determined only by how users perceive the quality of service. By both objective analysis and subjective assessment the relationship between network performance and Quality of Service is investigated for a high quality video/audio codec implementation.