Optimal capacity management of virtual paths in ATM networks

In this paper we address the problem of the virtual path (VP) bandwidth control in telecommunication networks implemented with ATM technique. The use of VPs allows a reduction of the processing load generated in the nodes during call access, set-up and routing operations, balanced by a waste of bandwidth that can be particularly relevant if VP capacity is not changed according to the offered traffic. A better transmission efficiency can be obtained resorting to some simple mechanisms of adaptation. We propose a control strategy by which the VP capacity is checked and eventually changed with a period t_u , in order to minimize a weighted sum of resource and control costs, while meeting a blocking probability constraint. The envisaged control system can be described by the triple (t_u, M, h), where M is bandwidth modularity and h is a threshold introduced in the system to limit continuous bandwidth adaptation. Two problems have been analyzed concerning: 1) the determination of the algorithm to assign VP capacity in the decision instants, which applies a forecast model and a decision table depending only on VP occupancy; 2) the selection of optimal values for control system parameters, which can be made off line by an optimisation module. For small-medium traffic size numerical results show that VP capacity adaptation can save up to 20%. The results confirm the effectiveness of the method, also considering the simplicity of the capacity assignment process