Models of adaptive load allocation in a distributed electronic digital exchange

The "linea UT" electronic exchanges are an example of a distributed system where the resource management problem is approached in a completely distributed way, without any centralized map for real-time operations. The aim of the paper is the evaluation of performance enhancement which can be obtained by transforming the distributed allocation algorithm from invariant to adaptive. The approach uses learning automata to balance the load among the distributed modules without signalling overhead. A simulation model is described to compare, in a dynamic environment, the performances of the invariant and the adaptive approach. An analytical reference model is used to validate the simulation model in static conditions.