Optimal design of AQM controllers

Active queue management (AQM) algorithms have been proposed to leverage TCP protocol inherent congestion algorithms which usually fail to maintain the queue length to an average level under large or bursting workloads. The main task of AQM algorithms is to detect network congestion at its early stages and start drop packets before the congestion affects the network throughput in a high degree. In this paper, an new router controller is proposed based on optimal control theory and is compared to the most important AQM algorithms according to their functionality, characteristics and their parameters. Furthermore, illustrative simulation experiments are conducted to evaluate overall system performance queue length response over different number of flows at different time intervals. The proposed algorithm results significant improvement of performance of the network and faster respond to external loads. Key-Words: AQM, router models, network congestion, optimal control.