RAN Network Information Base (R-NIB)

The R-NIB maintains a soft-state database that is continuously updated via the E2 interface with latency measured in milliseconds. It does not simply store static configuration; it reflects the dynamic, transient state of the network.

• UE Context: Stores per-UE information including Radio Network Temporary Identifier (RNTI), serving cell, and active bearers.
• Cell State: Tracks real-time Physical Resource Block (PRB) utilization, active UE count, and antenna tilt.
• Topology: Maintains a graph of neighbor relations and gNB-DU to gNB-CU mappings.
• Mechanism: Uses a publish/subscribe model where xApps register for specific data updates rather than polling.

The R-NIB organizes data into distinct layers of abstraction to serve diverse consumer needs without exposing raw, unstructured telemetry. This prevents xApp developers from needing to parse vendor-specific data formats.

• Physical Layer: Raw RF metrics including Channel Quality Indicator (CQI), Reference Signal Received Power (RSRP), and Timing Advance (TA).
• MAC Layer: Scheduling decisions, Buffer Status Reports (BSR), and Hybrid ARQ (HARQ) state.
• RRC Layer: UE connection states (RRC_IDLE, RRC_CONNECTED), handover history, and capability information.
• Application Layer: Correlated QoE metrics such as video stall rates and TCP round-trip time.

The R-NIB implements optimistic concurrency control to manage simultaneous read/write operations from multiple xApps. Since the Near-RT RIC hosts numerous independent microservices, the database must prevent dirty reads and write conflicts.

• Snapshot Isolation: Each xApp query operates on a consistent point-in-time view of the network state.
• Write Arbitration: Conflicting control commands (e.g., two xApps requesting different handover thresholds for the same UE) are flagged for the Conflict Mitigation function.
• Data Versioning: Every record includes a timestamp and sequence number, enabling rollback of erroneous state injections.

Unlike a flat key-value store, the R-NIB inherently understands the hierarchical and graph-based structure of the RAN. It models the parent-child relationships between network functions and the peer relationships between cells.

• Graph Representation: Stores the RAN as a directed graph where nodes are NFs (O-CU-CP, O-DU, O-RU) and edges represent F1, E1, and Xn interfaces.
• Spatial Indexing: Geospatially indexes cells and UEs to enable location-based queries (e.g., 'find all UEs within 200m of cell sector A').
• Slice Awareness: Tags all data records with their associated Single Network Slice Selection Assistance Information (S-NSSAI) to enable per-slice monitoring.

The R-NIB is not limited to E2 data. It ingests enrichment information from the Non-RT RIC via the A1 interface to provide xApps with a broader operational context that raw radio telemetry cannot supply.

• UE Mobility Patterns: Historical trajectory predictions generated by rApps to inform predictive handover xApps.
• Weather and Event Data: External data feeds correlated with cell load to anticipate traffic surges.
• Spectrum Occupancy Maps: Long-term interference maps generated by the Non-RT RIC's offline analysis.
• ML Model Inference Results: Stores the output of deployed models as first-class data for other xApps to consume.

The R-NIB translates vendor-specific data structures from disparate O-RAN network functions into a standardized, unified information model. This is the critical enabler of multi-vendor interoperability.

• Normalization Engine: Converts proprietary KPM (Key Performance Measurement) formats into O-RAN Alliance-defined service models.
• Semantic Mapping: Ensures that 'DL PRB Utilization' from Vendor A maps to the identical semantic concept as 'Downlink Resource Usage' from Vendor B.
• API Exposure: Provides a single, consistent RESTful or gRPC API to xApps regardless of the underlying RAN hardware vendor mix.

A microservice-based application hosted on the Near-RT RIC that consumes data directly from the R-NIB and executes near-real-time control logic. xApps subscribe to specific data types—such as L2 buffer status or channel quality indicators—and write optimization commands back to the network.

• Primary consumer of R-NIB data for closed-loop control
• Examples: Massive MIMO Optimization, QoE Optimization, Inter-Cell Interference Coordination
• Must resolve conflicts with other xApps via the Conflict Mitigation framework

The standardized open interface between the Non-RT RIC and the Near-RT RIC used for policy-based guidance and enrichment information. The Non-RT RIC uses the A1 interface to inject long-term AI/ML model updates, enrichment data, and policy directives into the R-NIB for consumption by xApps.

• Carries declarative policies that influence R-NIB state interpretation
• Enables model lifecycle management across the RIC hierarchy
• Operates on a non-real-time timescale (seconds to minutes)

The infrastructure within the SMO and RIC that aggregates performance measurements and telemetry from network functions. This framework acts as the pre-processing pipeline for the R-NIB, filtering, normalizing, and distributing data streams to registered xApps and rApps.

• Handles FCAPS data via the O1 interface
• Ensures data quality and consistency before R-NIB ingestion
• Manages subscription-based data distribution to authorized consumers

A coordination mechanism within the RIC that detects and resolves contradictory control commands issued by multiple concurrently running xApps. The R-NIB provides the shared state view that allows the conflict mitigation function to identify when two xApps are issuing incompatible resource allocation commands.

• Prevents network instability from competing optimization loops
• Uses R-NIB as the single source of truth for current network state
• Essential for maintaining deterministic behavior in multi-xApp environments

The automated pipeline within the SMO and Non-RT RIC that manages the end-to-end lifecycle of AI models. This system uses historical data from the R-NIB for model training and validation, then deploys optimized inference models back to the Near-RT RIC.

• Manages data ingestion, training, validation, and deployment
• Leverages R-NIB historical snapshots for offline model development
• Includes Model Drift Detection to monitor inference accuracy degradation