Proposed New SD-WAN Architecture to Facilitate Dynamic Load Balancing

Emmanuel, Ikiomoye

Abstract

SD-WAN (Software Defined Wide Area Networking) is widely regarded as the future of networking as it integrates programmable features into its architecture. However, there is currently no all-encompassing solution for traffic monitoring policies in the control plane that accounts for the dynamic behaviour of applications transported in the SD-WAN architecture. In a traditional SD-WAN controller configuration, load balancing is distributed across different WANs according to predefined policies. Furthermore, existing SD-WAN controllers do not account for the effect of load balancing on end-to-end protocol performance and vice versa. To address this gap, this study proposes a new logical SD-WAN architecture to facilitate dynamic load balancing of end-to-end protocol operations in the SD-WAN architecture for improved resource utilisation. This is achieved through interaction between the Enhanced SD-WAN controller's traffic management decisions system and end-to-end protocol performance, which requires deep packet inspection of network data flows to detect signs of congestion. The solution involves developing a traffic monitor function which is responsible for analysing, classifying, and managing traffic. Additionally, an enhanced SD-WAN controller called “En-SDWAN controller” focuses on sending route updates to edge routers to change the route of traffic flows that are experiencing congestion is implemented. The performance of the developed Enhanced SD-WAN architecture was assessed in reference to the traditional SD-WAN across various scenarios, including regular traffic, link failure, and when network congestion is experienced. In the normal traffic conditions, the performance of Enhanced SD-WAN closely resembles that of traditional SD-WAN, owing to the absence of congestion within the network architecture. The average throughput performance of Enhanced SD-WAN stands at 50.07%, while traditional SD-WAN operates at 49.93%. However, during network link failures, Enhanced SD-WAN demonstrates quicker convergence and better throughput performance compared to traditional SD-WAN by 1.46%. Similarly, during instances of network congestion, Enhanced SD-WAN exhibits an average throughput of 11.21Mbps, surpassing traditional SD-WAN's average throughput of 7.02Mbps. The implementation of both the GNS3 and Python applications validates the results and design, demonstrating consistency throughout. This outcome underscores the improved flow control of packets facilitated by route updates and dynamic load sharing implemented within the Enhanced SD-WAN controller. The implementation of a new logical architecture has enhanced the performance and efficiency of traditional SD-WAN systems by adapting to the evolving nature of applications and offering better control over network traffic.

Citation

Emmanuel, I. (2024). Proposed New SD-WAN Architecture to Facilitate Dynamic Load Balancing. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Apr 23, 2024
• Publicly Available Date: May 26, 2024
• Keywords: SD-WAN, SDN, Network Automation
• Award Date: Apr 25, 2024