The O-CU is where the RAN is centralised and made programmable. It runs PDCP, RRC and SDAP, split into a control plane (CU-CP) and a user plane (CU-UP) that scale independently, serving many distributed units over F1 and connecting to the core and the Near-RT RIC. Non-real-time, software-defined, and the natural home for data-driven RAN control.
At the top of the disaggregated RAN, above the O-DUs. It runs the non-real-time layers — PDCP, RRC, SDAP — connects to the DUs over F1, to the 5G core over NG, and to the Near-RT RIC over E2. One CU can serve many DUs, which is why it is centralised.
O-RAN separates the control plane (CU-CP: RRC and PDCP-C) from the user plane (CU-UP: SDAP and PDCP-U), joined by the E1 interface. The benefit is that control and throughput scale and deploy independently — you can place user-plane capacity near the traffic and keep control centralised.
Increasingly, yes. Because one CU serves many DUs and must scale with load, it is well suited to containerised, cloud-native deployment on commodity compute — which is a large part of why O-RAN exists.
Tell us the specification, the constraint and the deadline.