PERFORMANCE IMPROVEMENT OF SOFTWARE-DEFINED NETWORKING (SDN) BY ADJUSTING THE CONGESTION CONTROL, MULTIPATH ROUTING AND BANDWIDTH ALLOCATION TECHNIQUES

• Main Author: Naning Hertiana, Sofia
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36490
• Institution: Institut Teknologi Bandung

In today's Internet network, the use of Transmission Control Protocol (TCP) as a congestion control technique and single path routing as a routing technique raise several problems not only low link utilization and throughput, but also unfair and unbalanced traffic distribution. These problems happen because TCP and single path routing on the Internet currently do not have overall network information. Therefore, there is a need for congestion control and routing techniques that are assisted by networks using global information. The technique is obtained from the future network framework of Software Defined Networking (SDN). Separation of control plane and data plane at SDN makes this network more flexible to be developed in accordance with the development of service application technology. In this study, a controller-based congestion control scheme was proposed, called PACEC (Path Associativity Centralized Explicit Congestion Control). In PACEC, the setting of the transmission rate fluctuations is based on the bandwidth availability and path delay. In this case, the SDN controller is programmed as a determinant of congestion control policies. The advantage of the proposed scheme is to use bandwidth information known to the controller globally so that network resources can be used optimally and network control decisions can be made more accurately. PACEC is a modification and improvement of classical router assisted congestion control algorithms. In conducting congestion control, the PACEC scheme is integrated with multipath routing and bandwidth allocation. By adjusting this, every flow that is allowed to enter the network can be passed on several of the best paths and guarantee allocation rate for it to meet the QoS requirements. As a performance indicator of the proposed algorithm are throughput, delay, fairness, smoothness, overhead, and complexity. The performance of the proposed algorithm is verified by a series of tests using simulation software Mininet and RYU controller. The simulation results show that the proposed algorithm scheme produces better performance, i.e. increased throughput up to 10%, fairness 9%, increased efficiency reaching 50% and decreasing overhead up to 80%. This shows the advantages of the proposed scheme. However, future work is still needed to reduce complexity parameter.