Improved loss calculations at an ATM multiplexer
IEEE/ACM Transactions on Networking (TON)
Random Early Blocking Routing in VP-Based ATM Networks
ICOIN '01 Proceedings of the The 15th International Conference on Information Networking
A Very Accurate Approximation for Cell Loss Ratio in ATM Networks
ICON '01 Proceedings of the 9th IEEE International Conference on Networks
LLR routing in homogeneous VP-based ATM networks
INFOCOM '95 Proceedings of the Fourteenth Annual Joint Conference of the IEEE Computer and Communication Societies (Vol. 2)-Volume - Volume 2
Equivalent capacity and its application to bandwidth allocation in high-speed networks
IEEE Journal on Selected Areas in Communications
Packet Loss Probability Approximation in High-Speed Networks Based on Self-Similar Traffic
CNSR '09 Proceedings of the 2009 Seventh Annual Communication Networks and Services Research Conference
Traffic distribution for end-to-end QoS routing with multicast multichannel services
The Journal of Supercomputing
Hi-index | 0.24 |
ATM as a high-speed cell switching technology can support multiple classes of traffic with different quality of service (QoS) requirements and diverse traffic characteristics. A main QoS requirement is the cell loss ratio (CLR). We need a real-time expression for the CLR calculation in ATM networks where the statistical multiplexing is an important factor. The existing analytical methods for the CLR estimation are mostly based on fluid-flow and stationary approximate models. In this paper, we first evaluate these methods against the results obtained through simulation. The simulation is done at the cell level that provides very accurate results with buffer size as a variant. It is shown that the CLR estimation based on existing analytical models are widely overestimated. We have, then, proposed three new approaches that yield significant improvement in the accuracy of the CLR approximation. First, we have found global correction coefficients to compensate for the error of the current analytical methods. Second, we have proposed a new upper bound based on exact modeling of system behavior in the finite buffer case. This is a novel approach that combines fluid-flow and stationary approximate models and outperforms all the previous ones. The accuracy of the proposed model is verified by simulation. Third, we have found a tight piece-wise linear approximation that can be calculated in real-time. We have studied application of these bounds in non-homogeneous as well as homogeneous cases.