Time-driven fluid simulation for high-speed networks

We propose to use time-driven fluid simulation to simulate high-speed networks. In this simulation scheme, network elements are modeled as fluid servers yet the traffic source can be of arbitrary type, including a discrete-event source and fluid source. We prove that for single class fluid, the discretization error is within a deterministic bound proportional to the discretization interval length and is not related to the network size. We show that fluid simulation offers the flexibility to trade accuracy for efficiency, and it is also easy to parallelize. These two features combined together enable us to tremendously speed up simulation.