A PREDICTIVE APPROACH FOR ABR TRAFFIC MANAGEMENT IN ATM NETWORKS

We propose a predictive mechanism for managing the ABR traffic in ATM networks. For ABR traffic management, the establishment of a feedback loop between the source/destination end systems, and the network lets the data sources know the state of the network, and utilize the availa ble network bandwidth more efficiently. However, because of the propag ation delay of sending back the resource management cells to the data sources, the network is either underutilized (when the network load is light), or congested (when the network load is heavy) by the excessiv e cells that enter the network during the propagation period. The prop osed predictive mechanism based on the Trellis diagram predicts the co ntents of the resource management cells and makes rate adjustments bef orehand. We also show how the predictive mechanism can be used as an ' 'add-on'' to the current congestion control algorithms such as EFCI, P RCA and EPRCA. A performance evaluation demonstrates that the predicti ve approach reduces excessive cells and increases network utilization. (C) 1998 Elsevier Science B.V.