NLAR: A New Approach to AQM
Xunli Fan, Feng Zheng, Lin Guan, Xingang Wang · 2010
The traditional adaptive Random Early Detection (RED) algorithm allows network to achieve high throughput and low average delay. However it uses a linear dropping probability function, which causes high jitter in the core router. To overcome the drawbacks of queue jitter in the traditional adaptive RED algorithm, this paper proposes a new Non-Linear Adaptive RED (NLAR) approach based on the Active Queue Management (AQM) scheme, which provides a non-linear adaptation to the dropping probability function of the adaptive RED. NLAR enables the gradient of the dropping probability to vary along with the deviation that is between the average queue length and the target queue length, which contributes to a more stable algorithm. Empirical simulation with various data analysis have demonstrated that the NLAR algorithm outperforms the adaptive RED algorithm in most scenarios.