Efficient Resource Allocation and Link Recovery Mechanism for Resilient SDN/NFV-based Communications

• Main Authors: Chia-Wei Huang, 黃佳威
• Other Authors: Chung-An Shen
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/8rm6z6

碩士 === 國立臺灣科技大學 === 電子工程系 === 107 === It is crucial to achieve high reliability and low latency concurrently for networked applications such as smart grid, data center, and intellengent factory. Software-Defined Networking (SDN) and Network Function Virtualization (NFV) are two novel network technologies that are proposed to break the obstacle of network architecture. SDN put the control plane to a central controller to control network traffic flexibly. NFV virtualize the traditional network function machine into software which installed in the specific machine and could plan network resource flexibly and dynamically. However, they has some bottlenecks. Firstly, in SDN, traditional link recovery mechanism will cost much time and the other faster mechanisms will occupy too many memories in switches. Therefore, to find a fast mechanism with low memory cost is very difficult. Second, in NFV, how to allocate virtualized resources to end users is also a difficult problem. Consequently, Many researches has proposed some algorithms to solve Virtual Network Embedding (VNE) problem. Nevertheless, the efficient algorithm to solve VNE problem is still not proposed. In our thesis, for the first bottleneck, we present a link recovery mechanism which enhances the reliability of the network, maintains low communication latency, and reduces memory utilizations of the switching devices by introducing Segment Routing. The results of our link recovery mechanism could achieve low recovery time and low memory utilization. In the second bottleneck, we propose an efficient VNE algorithm which uses node-rankng approach with global resources and adopts genetic algorithm path splitting method. We also use the link cost function to evaluate the network performance, and the results shows that our VNE algorithm outperforms other works.