Complexity-Aware Dynamic Resource Allocation in NFV-Based 5G Core Networks
Ehsan Sargolzaei, Mehdi Rasti, Siavash Khorsandi · IEEE Access · 2026
Network Function Virtualization (NFV) has emerged as a key enabler for efficient resource management in 5G core networks, supporting time-varying traffic and stringent Quality of Service (QoS) requirements. In our prior work, the ACNRA algorithm was developed as an optimization-based solution for dynamic NFV resource allocation, achieving near-optimal performance at the cost of high computational complexity, which limits its scalability in large-scale and real-time scenarios. To address this limitation, we propose HACNRA, a complexity-aware heuristic that preserves the original optimization model, objective function, and QoS constraints while eliminating the dominant computational bottleneck. Specifically, the binary optimization stage is replaced with a Hungarian-based assignment mechanism, reducing the complexity to polynomial time. The proposed approach is evaluated on two representative real-world backbone topologies, namely BtEurope and Geant and benchmarked against the optimal solution, static and dynamic ACNRA, and state-of-the-art heuristics (ARA and HCA), with all results averaged over 10 independent runs and reported using 95% confidence intervals. Simulation results show that HACNRA achieves near-optimal performance with an average optimality gap of 10.7%, improving resource cost by 3.20%–7.12% and energy consumption by 7.99%–11.23% over ARA and HCA, respectively, while satisfying QoS requirements. QoS performance is evaluated via end-to-end (E2E) delay constraints using the QoS satisfaction rate. HACNRA converges within 6–8 iterations with an average runtime of 11.8 seconds, whereas ACNRA requires up to 25 iterations and about 50 seconds under identical conditions, confirming a substantial runtime reduction. These results show that HACNRA effectively bridges the gap between optimization accuracy and computational scalability for large-scale, real-time NFV-based 5G core networks.