5G • 5G-Advanced (Rel-18)
Distributed NWDAF Architecture
Rel-18 Hierarchical Network Data Analytics Function
Splits the NWDAF into Model Training Logical Functions (MTLF) in central clouds and Analytics Logical Functions (AnLF) at edge data centers for sub-millisecond inference.
Technical Explanation
Centralized analytics engines cannot process real-time telemetry from thousands of base stations. Release 18 (TS 23.288) establishes the Distributed NWDAF: heavy AI model training runs in centralized clouds on the Model Training Logical Function (MTLF), while lightweight trained models are distributed to edge Analytics Logical Functions (AnLF) co-located with edge UPFs for sub-millisecond local inference.
Key Functions
- Decouples centralized heavy model training (MTLF) from localized real-time inference (AnLF)
- Delivers sub-millisecond analytics decisions directly at edge network breakout points
- Reduces core network backhaul telemetry transport volume by over 70%
- Enables localized closed-loop slice assurance and predictive user mobility routing
Specifications
3GPP TS 23.288, TS 29.520 (Rel-18)
Interfaces
Nnwdaf (SBI)Nnef
Related 5G Concepts
Want to memorize 5G concepts like this one?
Study it with SuperMemo SM-2 spaced repetition flashcards.