Fast simulation of background traffic through Fair Queueing networks

Large-scale network simulation is widely used to facilitate development, testing and validation of new and existing network technologies. To ensure a high-fidelity experimental environment, we often need to embed real devices and have the simulator running faster than real time. Since the generation and movement of background traffic in a network simulation represents so much of the workload, we develop here techniques for modeling background traffic through switches that use Fair Queueing scheduling. Our work is an extension of earlier efforts that assumed all switches use First-Come-First-Serve scheduling. It turns out the the scheduling policy has an important impact on the logic of the earlier technique, and on the performance it delivers. We describe the algorithm and give experimental results that show that like the earlier work, very significant acceleration of background traffic simulation is achieved.