Analysis and improvement on the robustness of AQM in DiffServ networks

RIO is the primary queue management mechanism proposed for assured forwarding in the DiffServ framework. Although RIO can generally provide bandwidth guarantee, its performance in terms of both delay and loss is sensitive to traffic level. In this paper, we demonstrate this sensitivity problem by simulation and present a qualitative explanation for its origin. We propose two adaptive algorithms to overcome this problem. Simulation results show that they can effectively improve the robustness of RIO under different and dynamic traffic, and provide stable and quantitative performance of delay or loss.