AMES: A framework for fair bandwidth sharing

Fair bandwidth sharing has been an active research area since the early days of networking. Today´s networks employ packet dropping as the primary mechanism for reacting to congestion, while transport layer protocols are designed to adapt their rates based on the observed packet losses. Due to the increasing transmission speed and a disproportionate decrease in the end-to-end delay, future networks are expected to have a high bandwidth-delay product as compared to the networks of today. The cost of a packet retransmission increases with the increase in the bandwidth-delay product and decrease in buffer sizes. Hence, there is a need to develop new techniques, for a future clean-slate network design, to achieve fair bandwidth sharing that do not primarily rely on packet dropping. To this end, this paper develops a framework for fair bandwidth sharing called Access Mechanism for Efficient Sharing (AMES), where every core router in the network employs link-specific queuing, round-robin scheduling, and reacts to congestion by restricting the adjacent routers from transmitting, rather than dropping packets. The core-routers do not maintain any per-flow information, nor perform any flow-specific operation. We demonstrate the link utilization and fairness characteristics of the AMES framework through extensive simulations and compare it to the Core Stateless Fair Queuing (CSFQ) architecture developed in the literature.