Performance analysis of reactive congestion control based upon queue length threshold values

Abstract

Reactive congestion control based upon two thresholds is analyzed. Under the congestion control treated here, a multiplexer or transit node in the network recognizes congestion when its queue length reaches some pre-determined threshold value, denoted by H. The congested node then explicitly informs the sources sharing the node of its congestion in some way to mark a forward congestion notification bit in the header of cells if provided, or to directly send control cells to those sources. The sources then stop the cell transmission according to the feedback information received. The relief of congestion is recognized by the node when its queue length decreases to some low-threshold, L, and informs the sources that the sources can again start the cell transmission. In this paper, we investigate this sort of congestion control by forming a queueing network, in which multiple sources and a single node are explicitly modeled. Through an analysis, we discuss the impact of those two thresholds on performance of the congestion control.