GA-based optimal dimensioning of three-level traffic shaper for statistical multiplexing in ATM networks

In this paper we discuss genetic algorithm (GA)- based optimal dimensioning of the three-level traffic shaper (tlts) for statistical multiplexing in atm networks. As the objective function for the optimal dimensioning, we consider three alternatives - the variance of the output cell service periods, the variance of the output cell rates, and the sum of squared differences of adjacent cell service periods. We perform simulations for optimal dimensioning of the tlts for statistical multiplexing of vbr video services, and we confirm that the objective function based on the variance of the output cell rates is the most suitable among the three alternatives. The framework of the optimal dimensioning procedure introduced in this paper can be easily extended to other optimal design problems in atm networks for which traditional gradient- based optimal dimensioning methods or random search techniques are not easily applicable due to the nonlinearities and complexities of the related functions and the discreteness of the involved parameters.