ARED: a novel adaptive congestion controller

Active queue management (AQM) is an effective method to provide an early notification of network congestion by proactively dropping or marking packets. In this paper, we propose a novel adaptive RED scheme in order to overcome the drawbacks of original RED gateway, which adjust the maximum drop ratio to keep the average queue length around the target value using gradient descent method based on discrete deterministic mathematical model of TCP/RED. Our adaptive RED scheme not only maintains its performance independent of traffic loads, but also converges to the target value faster than other enhanced RED algorithms. Moreover, simulation results show that our adaptive RED outperforms existing AQMs like PI and REM at least in the perspective of queue dynamics.