Fast Convergence of Congestion Control Algorithm

摘要

Traditional congestion control algorithms exhibit low convergence speed to equilibrium in high BDP (Bandwidth Delay Product) networks. The Fast Max-Min Kelly Control (FMKC) is a new and promising protocol that performs well especially in fairness convergence speed. FMKC utilizes packet loss to switch temporarily into a fairing mode and thereby improve the fairness convergence speed. In this paper, we investigate FMKC, prove its stability with homogeneous delay and also study the convergence behavior of FMKC. Extensive simulations show that FMKC offers faster convergence rate than EMKC. Simulations also show that FMKC preserves the good properties of EMKC, such as delay-independent stability and max-min fairness in steady state.