On the performance of a hybrid network traffic model

We study a hybrid network traffic model that combines a fluid-based analytical model using ordinary differential equations with the packet-oriented discrete-event simulation. The hybrid model is important to large-scale real-time network simulations, where the packet-level emulation traffic is handled by discrete events and the majority of the background traffic is described more efficiently as fluids. We present a simple performance analysis of this hybrid approach. We propose three techniques— namely, pointer caching, update dampening, and dynamic time stepping—in an implementation of the hybrid model. Experiments show that these techniques can significantly improve the performance of the fluid-based network simulation.