An extension to Norton's equivalent (Q583734)

A general framework is presented for obtaining equivalent state equations for general Markovian queueing networks. The techniques used are based on sample-path properties and cover the case when the system is not necessarily in equilibrium. As the motivation of the authors stems from optimal flow control problems in computer communication networks, the idea of Norton's equivalent is needed in proving structural results such as separation theorems. Therefore it is shown that Norton's equivalent is the conditional estimate of the intensities of associated point processes. This reinterpretation makes it possible to give the Norton's equivalent much more generality than it was possible with the usual engineering interpretation. For multi-class Markovian queueing networks, it is shown that a first-order equivalent system of an isolated system can be obtained via the conditional estimates of intensities of the arrival and departure processes to and from that station. Separation results for optimal control flow problems in queueing networks have also been obtained. Examples which illustrate these concepts are given. The results obtained here generalize those which require the ``product form'' networks.