Neural Network-based Congestion Control for ATM UBR Service

摘要

The UBR service of the ATM network supports a general delivering mode for those data with less delay sensitive and cell loss, and offers the best effort service. The EPD (Early Packet Discard) techniques basically employ a threshold to define the current traffic status and classify the cells into two classes, the first cell and the non-first cell. In this paper, the proposed method is based on the EPD technique and employs the results learned from the neural network to design a better and practical strategy to improve the traffic over the UBR service in ATM network. The main principle of the proposed method is to provide an adaptive threshold, which will become larger to accommodate both kinds of cells if the traffic is light, and which will become smaller to only accept the non-first cell if the traffic is heavy. Simulation results show that our scheme can significantly improve the TCP traffic over the UBR service.