Performance Evaluation of Active Queue Management in Fat Tree Architecture on Data Center Network

In the age of the rapid growth of Internet technology, people are connecting to the0020network continuously. Request is made in seconds, resulting in congestion in the data center network. This makes the researcher tried to make a breakthrough in order to reduce the network latency. One of many proposals is the use of the active queue management (AQM). Active queue management focuses on dropping the network packets inside a buffer to reduce network congestion. This method is capable to dropping the packets according to probabilistic manner. In this paper, we evaluate different AQM algorithms to the Data Center Network (DCN) using fat tree architecture. We conducted some experiments using NS-3 to evaluate the network performance of different AQM algorithms. The simulation is done using a research based tool for computer network design, NS- software on Ubuntu Operating Systems. NetAnim, the visualization and monitoring tool for network is used for the evaluation. Throughput, packet drop, and delay are measured and compared. Best throughput is gained when ARED AQM is used which is 498.5 kpbs. It is 37 kbps higher than NLRED and 59 kbps higher than those of RED. In terms of packet drop, the ARED shows the lowest packet drop of 2230 packet drop. ARED also has the lowest delay, 0.132 second, which is slightly different from NLRED with 0.133 second delay, whereas RED's delay is 0.177 second.