Speaker :- Vishal T. Mudnur
Abstract :- Decoupling approximation for the performance analysis of IEEE 802.11 DCF MAC is critical due to the problem of state space explosion associated with its exact DTMC model. In this work, we study the validity of the two widely used models of DCF MAC based on the decoupling approximation, namely fixed point analysis andmean field model. In the current literature, the validity of the two models is verified from the nature of the solution of the models i.e. unique fixed point for the fixed point analysis and globally stable ODEs for the mean field model. We show, using two examples, that the above conditions are not sufficient to validate the approximation or the applicability of the models to study DCF MAC. In fact, we note that the nature of the solutions fail to characterize the correlation in the attempt process and the approximation is poor in such cases. We conclude that the models cannot be validated based on the nature of the model solutions.
In this work, we focus on IEEE 802.11 DCF MAC with a finite number of back off stages and monotone non-decreasing backoff values. We show that the mean field ODEs are
globally stable when the backoff values are monotone non-decreasing. Further, we note that the nature of the solution of mean field ODEs is similar to the fixed point solution (based on the decoupling approximation). This permits us to study the decoupling approximation that is fundamental to both the analyses. Using simulations, we characterize the correlation in the attempt process (backoff engines) of the nodes and study the impact of backoff parameters (e.g., minimum average backoff duration, backoff multiplier, number of stages) to identify the regimes where the decoupling approximation is appropriate. Based on our observations, we propose two simple techniques to validate the decoupling approximation and the analyses.