Simulating ATM integrated-services networks

In integrated-services networks, both the arrival processes and the performance objectives can vary greatly from one packet stream to another. Either the network must be designed to meet the strictest performance requirements by every possible measure of performance, or traffic control algorithms are needed that differentiate traffic based on performance objectives, and adjust performance accordingly. There is therefore a need to evaluate and compare the performance of such traffic control algorithms including algorithms for admission control, transmission scheduling, and packet dropping. However, ATM integrated-services networks place great demands on simulation systems. This paper describes an efficient approach to the design of discrete-event simulations for these networks. It is used to evaluate several different traffic control approaches. Some sample results from the simulation are presented which compare these algorithms. It is shown that cost-based scheduling is the most effective of the scheduling algorithms simulated, but the priority token bank, which is simpler to implement, is close behind.