Chance-constrained stochastic programming for integrated services network management

The design, dimensioning and traffic management of an Asynchronous Transfer Mode (ATM) network may be modelled as a hierarchical planning problem at different time-scales–capacity provision/expansion with intervals of weeks or months and real-time call routing at the time scale of the connection, which is itself treated in hierarchical order of transmitting calls (sec), bursts (msec) or cells (μsec). Taking the network topology as given, a chance-constrained stochastic program for integrated services network dimensioning and traffic management is formulated using Hui's notion of effective bandwidth. This allows the removal of probabilistic constraints at the call level and leads to consideration of the network management problem as a linear deterministic multicommodity flow problem. To allow flexibility and easy problem formulation with various objectives and network specifications the decision support tool MODLER is employed. The model described can be used for a prototype software system for future network design and management.