| A key issue in ATM and in any network architecture design is resource management, i.e., how to make the best use of available resources. However, striving for high utilization of a resource without proper allocation may lead to long queuing delays, and overflow losses, resulting in a bad user-perceived quality of service and hence, performance degradation.; With increased acceptance of broadband Networks for data & multimedia communications, the traffic is becoming more and more burstier. It is well known that the traditional data model (Markovian Model) can't accurately reflect the actual data both in wireless systems and the underlying ATM backbone. The traffic correlation leads to increased network congestion while average utilization of the network could remain poor. These problems necessitate development of efficient congestion control and performance management technique.; We propose a congestion control framework for real-time fractal traffic that can reduce the cell loss probability and increase the utilization of the links. The basic idea is to control the injection of the burst of cells at UNI (source) in an arbitrary high speed network so that the QoS could be satisfied, keeping the cell/packet loss probability at a pre-determined minimal level. The proposed algorithm also exploits the simple fact that the users do not always fully utilize the bandwidth allocated to them. This novel model is based on self-similarity as it has important impact on performance management. We have incorporated the dynamic inherited nature of the traffic as a monitoring control for changing the parameters in the congestion control mechanism. We have also extended it to apply in wireless environment. |
