Experimental Investigation of CAC and Effective Bandwidth for Video and Data

With the availability of commercial ATM switches implemented with large buffers and strict priorities between different traffic classes, it becomes possible to study ATM traffic control methods as well as integration strategies in an experimental environment where both real time and non-real-time traffic are present. In this paper, the connection admission control (CAC) issue is investigated in the Expert ATM Testbed with real switches and as realistic traffic sources as possible. Two types of sources have been used, MPEG video sources which are artificially generated based on real MPEG traces, and data sources which are modelled as traditional on/off sources. The video traffic represents real-time traffic and is given priority over the data traffic. To complement and verify the hardware experiments, two simulation tools have been developed. The first tool evaluates some shortcommings in the hardware generation of the artificial MPEG sources while the second tool uses the real MPEG traces and thereby constitutes an evaluation of the way the artificial model is made. Results from both simulation tools are compared with the results of the hardware experiments. Since the last few years have shown solid progress in the development of a concept for effective bandwidth which reduces the complexity of the CAC problem to the same level as for circuit switched networks, this paper also compares effective bandwidth results obtained by three methods from the literature with the results obtained from the experiments.