Optimal allocation of ATM networks resources by using a dynamic programming approach

A fundamental problem in a connection oriented ATM network that supports a broad spectrum of resource consuming services is defining the optimal way for network resources allocation. The network resources are specified as total bandwidth and total buffer size of the switching nodes. We formulate the problem of defining optimal resources that can be assigned to a set of different end-to-end connections as a minimization of a special recursive function with a number of independent arguments. This function includes an objective function and a set of functions related to the constraints imposed on the QoS characteristics of the end-to-end ATM connection. The objective function corresponds to bandwidth and/or the total buffer amount in the sequence of the ATM switching nodes appropriate to a given end-to-end connection. The typical constraints are imposed on cell loss probability, cell mean delivery time and cell variance delivery time in the end-to-end connection. The problem is solved by applying a dynamic programming concept. The solution includes an analysis of current network resources and the QoS characteristics which pertain to each node-to-node link of the end-to-end ATM connection.