Analytic Performance Evaluation of the RED Algorithm for QoS in TCP/IP Networks

The Random Early Detection algorithm is considered as a promising algorithm for Differentiated Services to cope with feedback oriented traffic like TCP. In this paper, we model the salient features of TCP and RED and evaluate the model using a discrete-time analysis. The novelty of our investigation is the analytic approach to the composed model, i.e., TCP feedback traffic over a RED queue. Deriving not only mean values but also distributions for the performance measures, we obtain insights of the behaviour of TCP under RED.