Stability and Fairness Analysis of AIMD/RED System with Heterogeneous Delays

In this paper, we systematically study the stability of Additive Increase and Multiplicative Decrease (AIMD)/Random Early Detection (RED) system, considering heterogeneous flows and feedback delays. By applying the methods of Lyapunov functional and Lyapunov function with Lyapunov-Razumikhin condition to the fluid model of the generalized AIMD/RED system, we obtain sufficient conditions to guarantee local asymptotic stability of the system with heterogeneous feedback delays. Our study also reveals the relationship between the AIMD parameters and the average window size of competing AIMD flows. Consequently, the TCP-friendly condition is derived. Numerical results with Matlab and simulation results with NS-2 are given to validate the analytical results. The analysis and the stability conditions derived can be used as a guideline to set up the AIMD/RED system parameters in order to maintain network stability and integrity, and to enhance system performance.