We consider open or closed multiclass queueing networks populated by many types of jobs. Such networks can model complex manufacturing systems like wafer fabrication facilities and computer communications systems. It is known that under many common service policies, the network can be unstable in the sense that the work-in-process grows to infinity even though every server is underutilized.
In these two talks, I will discuss how a corresponding deterministic, continuous fluid model can be used to study the stability of a stochastic, discrete queueing network operating under a particular service policy. I will also discuss how insights from studying fluid models can be used to design ëfficient policies" for stochastic networks.