Introduction. Self-similar traffic in modern telecommunication radio networks requires new methods for calculating parameters of adaptive Carrier Sense Multiple Access (CSMA) protocols, different from those adopted in classical teletraffic theory based on Poisson distribution models. Purpose. Construction of a mathematical model of the operation of adaptive synchronous CSMA protocols in conditions of self-generating traffic for three strategies for changing the length of data packets, and obtaining equations for the average transmission rate and comparing the effectiveness of adaptive control with the results of studies of the same protocols for traffic with a Poisson distribution. Methods. This goal is achieved by creating and analyzing a mathematical model of the operation of the flexible and rigid synchronous adaptive CSMA protocol for three strategies for changing the length of data packets under traffic conditions with a Pareto distribution. The model is described by the average protocol transmission rate equations, which are a function of traffic intensity, Pareto distribution parameters, and strategies for changing the length of data packets. Results. It has been proven that self-similar traffic significantly reduces the stability limit of adaptive CSMA protocols, which must be taken into account when using these protocols in radio networks with long-term dependence in traffic distribution. At the same time, the throughput of adaptive CSMA protocols remains virtually unchanged compared to the traffic model with a Poisson distribution. Conclusions. The proposed mathematical model of adaptive CSMA protocols allows to calculate the real values of the average speed, stability limit and throughput of these protocols under conditions of self-similar traffic for the proposed strategies for changing the length of data packets. The considered strategies (except for the second) are effective for adaptive control in conditions of self-generating traffic, but at a significantly lower value of traffic intensity compared to the Poisson distribution.
Read full abstract