Accelerating ATM Simulations Using Dynamic Component Substitution (DCS)

Abstract

The steady growth in the multifaceted use of broadband asynchronous transfer mode (ATM) networks for time-critical applications has significantly increased the demands on the quality of service (QoS) provided by the networks. Satisfying these demands requires the networks to be carefully engineered based on inferences drawn from detailed analysis of various scenarios. Analysis of networks is often performed through computer-based simulations. Simulation-based analysis of the networks, including nonquiescent or rare conditions, must be conducted using high-fidelity, high-resolution models that reflect the size and complexity of the network to ensure that crucial scalability issues do not dominate. However, such simulations are time-consuming because significant time is spent in driving the models to the desired scenarios. In an endeavor to address the issues associated with the aforementioned bottleneck, this article proposes a novel, multiresolution modeling-based methodology called dynamic component substitution (DCS). DCS is used to dynamically (i.e., during simulation) change the resolution of the model, which enables more optimal trade-offs between different parameters such as observability, fidelity, and simulation overheads, thereby reducing the total time for simulation.The article presents the issues involved in applying DCS in parallel simulations of ATM networks. An empirical evaluation of the proposed approach is also presented.The experiments indicate that DCS can significantly accelerate the simulation of ATM networks without affecting the overall accuracy of the simulation results.