Supporting Network Management with Real-Time Traffic Models

Real-time routing and flow control in circuit-switched networks is investigated. An algorithm which updates routing tables and flow-control parameters according to changing load conditions is derived. The network is described by means of stochastic difference equations. A control structure imposed by hardware requirements and realistic network status information patterns is considered. It is shown that the global objectives can be achieved by means of shortest-route algorithms with state-dependent route lengths. Implementation issues which are related to traffic estimation and prediction are discussed. The performance of a particular algorithm implementation is investigated by simulation. >