Stochastic Modeling and Analysis of GRED-I Congestion Control for Differentiated Bursty Traffic

This paper develops an analytical performance model for a finite capacity queueing system with an enhanced random early detection (RED) congestion control scheme in the presence of differentiated classes of bursty traffic. The aggregate traffic is captured by the superposition of 2-state Markov modulated Poisson processes (MMPP). Two individual thresholds are assigned to each traffic class in order to differentially control traffic injection rate. The accuracy of this model is verified by comparing the analytical results against those obtained from simulation. The model is adopted to investigate the effects of traffic burstiness and system capacity on the performance of the congestion control scheme.