Explicit rate congestion control with binary notifications

Abstract

This paper presents a new rate-based source adaptation algorithm for packet-switching networks. The proposed scheme relies on binary notifications from the network to avoid congestion. The network nodes are supposed to estimate a fair share of source rate either based on a stored per-flow information or using a stateless approach. Binary notifications are sent to the sources, reflecting a positive or negative difference between the source rate and the estimated fair share at the nodes. Based on these notifications, sources increase/decrease the current rates using an adaptive algorithm that performs better than the classical additive increase multiplicative decrease strategy. In particular, oscillations on the source rates are much smaller, even with large network delays. Simulation results are presented both for ATM ABR and Internet environments.