Statistical adapting RED in dynamic networks

Active Queue Management (AQM) aims to provide high link utilization and low queuing delay in communication networks. However, it is challenging to adapt AQM parameters in response to dynamic network scenarios with varying round-trip time, link capacity and traffic load. In order to maintain a stable queue with desired performance such as high link utilization and low queuing delay, this paper proposes a Statistical Adapting RED (SA-RED) to dynamically tune RED parameters based on the standard deviation of instantaneous queue size, which can be readily measured in practice. The proposed mechanism and corresponding algorithms can be implemented easily because they avoid the problem of measuring network parameters, such as the round-trip time and number of TCP flows. In addition, they are effective in achieving high performance, i.e., maintaining low queuing delay while providing the desired transient and steady-state performance under widely varying network conditions. These advantages are demonstrated by extensive NS-2 simulation experiments.