Evaluation of the Available Bit Rate Service Category in ATM Networks

In this paper we evaluate various aspects of the ABR flow control schemes using three different evaluation techniques: analysis, simulation and experiments. A discrete-time Markov model alows the evaluation of the influence of different system parameters (distance between source and switch, burstiness of CBR/VBR traffic, time in between computation of Allowed Cell Rate and burstiness of ABR traffic)) on the ABR traffic throughput and the ABR buffer requirements. This study leads to some simple engineering rules for the ABR service category. We investigate through simulation the ability of the ERICA+ switch algorithm to provide minimum throughput guarantees to Internet TCP/IP traffic. In addition, tow types of experiments have been performed, each one using different ABR equipment. The first set of experiments have been carried out at the EXPERT testbed in Basel (CH) using one of the pioneer commercial ABR switches by ATM Systems. In the experiments we investigate the efficiency of the ABR service when the network resources are shared with guaranteed rate traffic and the effect of a long distance delay on the feedback loop. The second experimental investigation is based on an ABR Access Service Multiplexer, developed within the Corporate Research Center of Alcatel, Antwerp. In this case, the performance and buffer requirements of TCP traffic multiplexed with priority VBR background traffic in an ABR queue is evaluated. The outcome of this study is a set of guidelines for dimensioning an ABR access configuration for the efficient transport of Internet TCP traffic.