MCA: an end-to-end multicast congestion avoidance scheme with feedback suppression

Abstract

In this paper, we propose MCA, an end-to-end multicast congestion avoidance scheme with feedback suppression. Congestion avoidance [J. Comput. Netw. ISDN 17 (1989) 1] is different from congestion control in the sense that our scheme detects and responds to network congestion without necessarily inducing packet loss . Our scheme is a single-rate scheme and operates end-to-end, i.e. the sending rate is controlled by the source based on feedback from the most congested receiver and does not expect packet marking or other support from intermediate nodes . We design it to be robust under both lossless and lossy situations. Congestion is detected at receivers using the concept of accumulation (the number of buffered bits of a flow inside the network) and simple thresholding techniques proposed in our recent unicast work [Accumulation-based congestion control (submitted)]. For the purposes of choosing representative (the most congested receiver) by the source and suppressing feedback by receivers, each receiver maintains its Good Throughput Rate At Congestion (G-TRAC) (the product of receiving rate during congestion epochs and 1− f , where f is congestion occurrence frequency).In this way, receivers do not need to continuously (either densely or sparsely) exchange packets with the sender (e.g. to measure RTT). Therefore, MCA is scalable for large-size groups. We evaluate the design and demonstrate the performance of MCA using detailed ns -2 simulations.