Hybrid discrete-continuous fluid-flow simulation

The fluid-flow paradigm is an alternative approach to the traditional queueing paradigm, where workload movement in a network is described in terms of a fluid metaphor rather than the familiar customer metaphor. Stochastic fluid network models can considerably speed up the simulation of complex flow networks, such as high-speed packet-based telecommunications networks, where traditional queueing-based simulation routinely gives rise to enormous numbers of events. Furthermore, they are also amenable to fast, nonparametric, and unbiased estimation (unlike traditional queueing) of IPA (Infinitesimal Perturbation Analysis) gradients (derivatives of performance measures with respect to various parameters), which may be used to make decisions on network design, replenishment, management or control. This paper outlines the design of a fluid-oriented hybrid simulator admitting both packet flows and fluid flows. It focuses on the fluid processing design and on efficiency issues.