Virtual path system optimization in ATM networks with stochastic traffic flows

The traffic control is a critical issue in ATM networks. In ATM, the traffic control is implemented at different levels: cell level, call level and network flow level. The virtual path (VP) distribution involves both call level and flow level controls. The VP distribution is a logic network design problem based on the physical network. Several VP optimization schemes have been proposed, and a large number of these schemes are based on the flow assignment (FA) model. In this paper, an improved flow model is proposed with a non-linear objective function. The proposed model incorporates two concepts: VP capacity and VP flow, to perform the optimization. The proposed model distributes traffic on all available VPs evenly, and has redundant capacities. Hence, the resulting VP system is resilient to input traffic changes and physical link failures. In addition to the proposed FA model, we introduce a stochastic programming (SP) methodology to allocate virtual paths when the incoming traffic changes stochastically. Experimental results show that the proposed flow model and the stochastic methodology improve the performance of ATM networks.