FlexScale: Scalable and Efficient Management Approach for Near-RT RIC in O-RAN

The Near-Real-Time (Near-RT) Radio Access Network (RAN) Intelligent Controller (RIC) in the Open RAN (O-RAN) architecture provides flexibility and programmability, enabling dynamic network management through Machine Learning based applications known as xApps. However, scalability and strict latency requirements hinder the support of numerous xApps. Existing orchestration solutions struggle to efficiently manage large-scale xApp deployments while maintaining latency below one second. Connecting multiple next-generation NodeBs (gNBs) to a single Near-RT RIC risks performance bottlenecks and single points of failure in the O-RAN architecture. To address these challenges, we propose FlexScale, a scalable approach to enhance O-RAN system capacity for extensive xApp deployments. It dynamically scales Near-RT RIC instances connected to gNB modules (E2 Nodes) using Kubernetes-based Horizontal and Vertical Pod Autoscaling (HPA, VPA) and native load balancing. FlexScale optimizes CPU utilization while meeting latency requirements. Simulations show that FlexScale efficiently scales xApps across multiple Near-RT RIC pods, addressing system limitations. Under high E2AP traffic, HPA and VPA autoscaling reduce latency by up to 96% compared to a single Near-RT RIC deployment. Additionally, CPU usage is reduced by approximately 70%, ensuring balanced resource utilization during traffic fluctuations. FlexScale demonstrates its capability to support large-scale xApp deployments while maintaining performance and efficiency.