TAHA: Traffic-Aware Hybrid Auto-Scaling of VNF Resources for 5G/B5G
Kanchan K. Tiwari, Krishna Moorthy Sivalingam · 2025
This paper studies auto-scaling of network function instances in a 5G Core network system based on Network Function Virtualization (NFV). Auto-scaling techniques have become essential for handling diverse traffic types while maintaining high Quality of Service (QoS). Scaling refers to the dynamic process of adjusting computational resources, either by increasing or decreasing them, to accommodate varying traffic loads. This paper introduces a hybrid reactive and proactive scaling mechanism. Proactive scaling minimizes Service Level Agreement (SLA) violations for infrastructure providers (IPs), while reactive scaling acts as a safeguard against forecast errors. We evaluate the proposed technique using the discrete event simulator SimPy, and Mininet emulation. Extensive numerical studies under dynamic traffic conditions modeled by a finite Markov chain demonstrate that our approach achieves a latency reduction of 92 % compared to the baseline non-scaling system, and at least$\mathbf{4 6 \%}$reduction compared to existing schemes.