Achieving max-min fairness by decentralization for the ABR traffic controlin ATM networks

The available bit rate (ABR) is an ATM service category that provides an ec onomical support of connections having vague requirements. An ABR session m ay specify its peak cell rate (PCR) and minimum cell rate (NICR), and avail able bandwidth is allocated to competing sessions based on the max-min poli cy. In this paper. we investigate the ABR traffic control from a different point of view: Based on tho decentralized bandwidth allocation model studie d in [9], we prove that the max-min rate vector is the equilibrium of a cer tain system of noncooperative optimizations. This interpretation suggests a new framework for ABR traffic control that allows the max-mm optimality to be achieved and maintained by end-systems, and not by network switches. Mo reover, in the discussion, we consider the constrained version of max-min f airness and develop an efficient algorithm with theoretical justification t o determine the optimal rate vector.