Approximative Markov Control of Multiservice Telecommunication Links

摘要

A modern circuit-switched telecommunication network offers point-to-pointconnections. The problem in network control is to choose the links along which each connection is routed, and to decide which connection requests should be denied in order not to exceed the network capacity. With certain assumptions these decisions can be made optimally by the methods of Markov decision theory; however in practice the Markov process model of a single link can contain as many as 10 to the 20th power states, and the known exact algorithms for finding the optimal control policy require as many variables as there are system states. The work contains the necessary background on Markov decision theory in the case of a finite state space and the infinite horizon average cost criterion, as well as the basics of teletraffic theory beginning with the traditional Erlang link model for telephone networks, and proceeding to the generalization of the model to modern telecommunication networks where the carried connections can be divided into traffic classes of different characteristics.