Diffusion Approximation Model of TCP NewReno Congestion Control Mechanism

Abstract

Abstract The aim of the article is to contribute to the modeling and evaluation of TCP/IP traffic control in case of active queue management mechanisms implemented in congestion router. We consider mechanisms based on RED and fractional-order $$PI^\alpha$$PIα controller. A new model based on diffusion approximation is developed and its results are compared with the results of an existing fluid flow model. Both models are used to investigate the influence of AQM control parameters on the flow control. Their numerical results concerning dynamics of congestion widow, loss probability, and traffic changes are quite similar. The advantage of the new model lies in giving more detailed information: the distribution of router queue and the distribution of introduced delay, not only their mean values as it is in case of fluid flow model. The variability of transmission time (jitter) is a factor of transmission quality, and therefore, the proposed model may be useful in further studies on TCP/IP quality of service.