An Active Queue Management Mechanism for Minimizing Queueing Delay

Real-time Internet applications, such as online gaming, VoIP (Voice over Internet Protocol) and teleconferencing, prefer short latency over high throughput and have been steadily rising in popularity as of 2021. Yet current Active Queue Management (AQM) algorithms inherently create delay-inducing standing queue, thus cannot effectively minimize the latency. We propose an adaptive mechanism named MID (Multiplicative Increase/Decrease) in this paper, to minimize queueing delay by eliminating the standing queue created by AQMs relying on fixed targets, based on the assumption that standing queue can be removed without substantially affecting the network performance in other aspects. The proposed mechanism discovers the optimal control target by gently reducing the target as long as the buffer is not empty, otherwise the target will be increased and the discovery process will be temporarily suspended. All parameters of MID are insensitive to network environments hence require no tuning. The paper validates the performance of the proposed algorithm by applying it to PI and PIE in simulations, and compares the performance between the MID-enhanced AQMs and their original counterparts. Results confirm that MID is capable of efficiently minimizing queueing delay at an acceptable cost.