BRIDGING TCP AND ABR CONGESTION CONTROL ALGORITHMS

Abstract

The Asynchronous Transfer Mode (ATM) networks are quickly being adopted as backbones over various parts of Internet. In this thesis we studied and analyzed the performance of TCP/IP protocol over ATM network's Available Bit Rate (ABR) service. We also studied the negative interaction between TCP (Transmission Control Protocol) and ATM-ABR congestion control algorithms. This negative interaction causes the poor performance of TCP/IP over ATM-ABR service. In order to reduce the negative interaction between two congestion control algorithms it is necessary to make an agent that bridges the gap between the TCP and ATM layers. Through the simulation we show that the Agent Based Algorithm adjusts the congestion window (CWND) of TCP source so that there is less cell loss in the network. The congestion window is partially controlled by the Agent. The introduction of agent at the edge of the network coordinates the TCP and ATM sending transmission and clearly indicates the strong correlation coefficient between CWND of TCP and actual cell rate (ACR) of ATM layers. By introducing early feed back congestion from the network there is reasonable improvement in the performance. The simulation results show that coordination at the edge of network and by early feed back about congestion from the network makes better use of available resources and improves the performance of TCP over ATM for different performance metrics such as Cell Loss Ratio, Packet Retransmission Ratio, and Throughput.