LRURC: A Low Complexity and Approximate Fair Active Queue Management Algorithm for Choking Non-Adaptive Flows

Abstract

Active Queue Management is an effective way to improve the TCP's performance. However, due to the existence of non-adaptive flows, most previous schemes cannot achieve efficiency, fairness, and easy-to-deploy simultaneously. In this paper, a novel scheme, named LRURC, is proposed to improve the fairness among different flows and reduce the flow completion time when the high-bandwidth non-adaptive flows arise. Motivated by the CHOKe algorithm, the LRURC adopts the sample-match mechanism to identify high-bandwidth flows. Furthermore, the least-recently-used mechanism is introduced to manage the high-bandwidth flows. Different from most other schemes, the LRURC calculates the packet dropping probability according to the flows' rate, which is estimated by the sample-match mechanism. This can achieve fairness effectively and reduce the influence of high-bandwidth non-adaptive flows. When the flow is not hit by the sample-match mechanism, the packet dropping probability remains constant. Simulation results show that LRURC can improve the fairness among different flows, reduce the completion time of adaptive flows, and hold the queue length around an expected value when high-bandwidth non-adaptive flows arise.