ATM網路逐個鏈結式預測之流量控制

Abstract

在這篇論文中，我們提出一種在即時流量的同時存在下，傳輸可控制的流 量（controlled traffic）的預測式流量控制方法。這些即時流量擁有嚴 格的可延遲要求，如聲音、視訊的流量。這個控制方法嘗試去預測在未來 的時段，下游節點的緩衝區使用量與cross traffic到達的速率並動態調 整上游節點的傳送速率。我們的控制方法是根據下游節點傳回的資訊作調 整。我們針對多樣的cross traffic參數、傳輸距離、回授間隔等對我們 的控制方法的影響作研究。我們也和其他的控制方法比較，我們的發現顯 示預測式的控制表現遠比靜態控制或適應性末端對末端的控制來得好。 In this thesis, we present a predictive flow control scheme for carrying controlled traffic in the presence of real time traffic (cross traffic) which has stringent delay requirements such as voice and video traffic. The control scheme attempts to predict the buffer content and cross traffic arrival rate of downstream node at some time in the future and adjusts the sending rate of upstream dynamically. The policy is based on information feedback from downstream. We study the performance of our scheme over a wide range of cross traffic parameter values, distance and feedback interval. We also compare the control scheme to other policies. Our investigation shows that predictive control scheme performs significantly better than static or adaptive end-to-end rate control policies.