Availability Guaranteed and Resource Efficient VNF Placement in SDN/NFV-Enabled Network through Traffic Forecasting
Yi Yue, Bo Hao Cheng, Shiding Sun, Wencong Yang, Xiongyan Tang · 2025
Network Function Virtualization (NFV) enables the realization of dedicated, proprietary network functions as software, which we can instantiate flexibly on commodity servers as Virtual Network Functions (VNFs). This approach facilitates significant cost reduction and operational flexibility. However, NFV also introduces new challenges, particularly regarding the availability of network services during the VNF deployment process, due to the inherently error-prone nature of software. The issue of ensuring high availability in VNF deployment has garnered considerable attention in the academic community, with redundancy provisioning commonly regarded as the standard solution. Additionally, the time-varying traffic in operator networks complicates the deployment process. Accurate traffic prediction enables operators to dynamically scale VNF instances based on demand, optimizing resource usage and reducing costs. Building on these considerations, we investigate the availabilityaware VNF deployment problem within data center networks. We incorporate a redundancy-sharing mechanism alongside traffic forecasting method to enhance resource utilization efficiency. We formally model the problem and propose an Availabilityguaranteed and Resource-efficient VNF Placement (ARVP) for mapping Service Function Chain Requests (SFCRs) in SDN/NFVenabled networks. We conduct a comprehensive numerical simulation to evaluate the performance of our proposed approach, comparing it against four alternative schemes from the existing literature. The results demonstrate that our algorithm outperforms the benchmarks regarding SFCR acceptance rate and activated nodes. Furthermore, it achieves up to 55 % resource savings when the availability requirement is six nines ($\mathbf{0. 9 9 9 9 9 9}$).