Markov Based Computational Model for Performance Evaluation of Congestion Control Variants

Analytic modelling of TCP congestion control has been a subject of intense research in the past two decades for both wireline and wireless high-speed Broadband networks. There has not been a single simple analytical approach that covers a wide range of TCP congestion control variants, making it hard to compare them against each other.. The contribution of this paper is generalization of a Markov model, previously used for TCP Cubic, so that it can also be used for TCP NewReno and a wide range of random loss rates. We also argue that the model can be easily extended to other mechanisms, even model-based mechanisms and those that are hard to define mathematically. We show that the results of the model tally with literature results. The model captures correctly the behaviour of TCP NewReno and TCP Cubic in environments with high random loss rates. We further confirm the results with NS3 simulation.