STATE-DEPENDENT ROUTING ON SYMMETRICAL LOSS NETWORKS WITH TRUNK RESERVATIONS .1.

We investigate a distributed, state-dependent, dynamic routing strateg y, called here aggregated-least-busy-alternative (ALBA), for circuit-s witched loss networks. The networks considered are symmetric and fully connected, the offered calls form Poisson streams, and routes have at most two links. In ALBA (K), the states of each link are lumped into K (K greater-than-or-equal-to 2) aggregates, and the route of each cal l is determined by local information on the aggregate states of the li nks of the alternate routes at the time of that call's arrival. The la st aggregate is always the set of states reserved for direct traffic. We give a fixed-point model for ALBA(K) for general K. The particular case of ALBA in which there is no aggregation is least-busy-alternativ e (LBA); ALBA(2) represents the other extreme of aggregation. We compa re simulation and analytic results for LBA, and find the agreement sur prisingly good. The structure of solutions computed here for LBA impli es that with just a few properly designed aggregates, it is possible t o closely approximate the solution (and thus the performance) of LBA. This is confirmed by results presented here on the comparative perform ances of LBA and ALBA(K), K = 2,3,4. We also consider an asymptotic sc aling based on the fixed-point models. It is shown that there is a dic hotomy in network behavior: if the offered traffic is below a threshol d, then the network loss probability decreases exponentially with incr easing network size, and above the threshold, performance is poor. Hen ce, the threshold sharply delineates ''engineered'' designs.