Performance Evaluation of Active Queue Management Algorithms in Large Network

Abstract

The large network architecture of today consists of thousands of computers connected together through many interconnecting router and switches. As many of them communicate concurrently, congestion over the channels may increase. In order to maintain the fairness among the transmission control protocol (TCP) based packets being sent, during congestion each flow of communication must be treated fairly. The most popular scheduling scheme implemented today is the drop-tail, which it cannot guarantee fairness and as the queues get full; the latency increases. Therefore, packet drops cause reduction in throughput of those flows. Research is ongoing to find an Active Queue Management (AQM) scheme that can solve this fairness issue, though fairness is not the only important feature that is looked on. Latency is another feature, that is important as well as the packet loss ratio. In this paper, the performance of network metrics between the drop-tail and the behavior of different Active queue management schemes (PFIFO, RED, ARED, CoDel) in large networks is evaluated using the open source discrete network simulator (NS-3). The results show that the throughput of CoDel and RED is the highest with the minimum latency. Moreover, the simulation results show also that Adaptive Active Queue Management (AAQM) provides postponing congestion and high link utilization whilst maintaining packet loss. The proposed scenario could be extended to evaluate real large network. The results prove the RED and CoDel have the best throughput and less latency comparing to others. Moreover, the ARED presents better performance when it compared to RED.