AI-Driven Resource Allocation and Auto-Scaling of VNFs in Edge-5G-IoT Ecosystems
Rafael Moreno‐Vozmediano, Eduardo Huedo, Ruben Santiago Montero, Ignacio Martin Llorente · Electronics · 2025
With the rapid expansion of edge-5G-IoT ecosystems, the need for intelligent and adaptive resource management strategies has become a critical challenge. In these environments, Virtualized Network Functions (VNFs) deployed at the network edge must handle highly dynamic workloads, making fixed resource allocation inefficient. While over-provisioning can lead to unnecessary resource waste, an especially critical issue in edge environments with limited resources, under-provisioning can degrade performance and service quality. This paper presents an AI-based predictive auto-scaling framework designed to optimize resource allocation for VNFs in edge/5G-enabled IoT environments. The proposed approach evaluates and integrates different ML-based regression models to characterize VNF resource consumption, along with various forecasting methods to anticipate future workload fluctuations, enabling both vertical and horizontal auto-scaling. Extensive experiments with real-world traffic data demonstrate the effectiveness of our approach, showing significant improvements in resource efficiency compared to fixed allocation strategies.