[B-6] Network Systems

2:15 PM - 2:30 PM JST(5:15 AM - 5:30 AM UTC)

[B-6-73] Cross-Layer Scaling Optimization for 5G Core on Public Cloud Using Control Plane Signaling

〇Katsuma Miyamoto¹, Koichiro Kunitomo¹, Kaito Shibutani¹ (1. NTT DOCOMO)

Keywords:

5G Core Network,Public Cloud,Auto-Scaling,Kubernetes

Public cloud deployment of 5G Core networks (5GC) faces a critical challenge in cost optimization under pay-as-you-go pricing models. This paper proposes a signal-driven cross-layer scaling optimization method that leverages 5GC control plane signals as demand indicators to jointly control the Pod and VM layers. Specifically, Initial Registration is utilized as a leading indicator for predictive VM-layer scaling, while Service Request and Handover are used as reactive indicators for Pod-layer scaling. Simulation evaluation at a scale of one million UEs demonstrates that the proposed method achieves approximately 37% reduction in VM running time compared to a conventional CPU-utilization-based approach, while maintaining zero SLA violations.