Architecture
How OCUDU is decomposed into CU-CP, CU-UP, DU-High, DU-Low, and RU, the standardized interfaces between them, and the runtime model that lets the same binaries run co-located or split across machines.
RAN & O-RAN Overview
Documentation
Specifications, reference implementations, deployment guides, and contribution workflows for OCUDU India.
Sections
- Architecture upper-PHY layering, interfaces, platform assumptions.
- Hardware Acceleration offload CPU-heavy upper-PHY stages to vRAN accelerators. Filled in.
- Deployment reference blueprints, hardware profiles, network topologies.
- Contributing how to file issues, propose RFCs, and submit pull requests.
Trace Studio
A real packet capture from a running OCUDU gNB, reconstructed as a unified timeline across UE, DU, CU-CP, and AMF. 63 control-plane messages, two PDU sessions, full lifecycle.
FAPI Split
Disaggregated OCUDU deployments - splitting the gNB into independent L1 and L2 processes across hosts or DPDK domains via the xFAPI translator-bridge.
Multi-UE Simulation
Run many OAI UEs against a single OCUDU RAN gNB over ZMQ, with a software IQ-superposition proxy, per-UE Linux network namespaces, a per-UE traffic engine, and a terminal dashboard for control and monitoring.
End-to-End Network Slicing
A complete open-source 5G network-slicing stack around OCUDU: multi-UE OAI on independent S-NSSAIs, slice-aware DU scheduling and CU admission/mobility, a multi-slice 5G Core, and standard O-RAN O1 management with closed-loop slice assurance through an SMO.
Hardware Acceleration
Offload CPU-heavy upper-PHY stages to vRAN accelerator cards for lower latency, higher throughput, and reduced host CPU.
ML-Based UL MCS Link Adaptation
A gradient-boosted model picks the uplink MCS by predicting decode probability per candidate, with an OLLA fallback and a self-adapting online-training sidecar. No ML runtime is linked into the RAN.
Deployment
Reference blueprints for real deployments - from a single edge site to a multi-site operator footprint.