CNF deployment of O-RAN Components
Scope
This page discusses the requirements, concepts and ideas about the deployment options and initial configuration of O-RAN Components in different cloud instances.
O-RAN Components are:
Near-RT-RIC (near real time RAN intelligent controller)
O-CU-UP (O-RAN centralized unit - user plane)
O-CU-CP (O-RAN centralized unit - control plane)
O-DU (O-RAN distributed unit)
According to the O-RAN architecture such components needs to be instantiated in different cloud location creating the RAN. In addition several interfaces between the O-RAN components, between O-RAN components and the Service Management and Orchestration Framework (SMO) and between O-RAN components and the Mobile core must be instantiated, configured and functionality tested, before the RAN is functional end-to-end.
The bases of such functionally was made in ONAP Frankfurt release and demonstrated several times.
Requirements
Pulling CNF images/containers from a centralized catalog
@Konrad Bańka - Currently it's not possible to include docker images (or others) inside NS/xNF Model, however as long as image is downloadable from registry accessible by cloud region (for example: LF nexus), there is no problem to run it
Instantiation of a CNF in an cloud environment
same environment as the SMO/ONAP is running (done)
different cloud locations (regional clouds)
Initial configuration of the CNF including
Security settings - @Konrad Bańka - could be managed with CDS Day0/1 config - see below
Certificates
Access control
Interface end points for topological neighbors (A1, O1, E2, F1, N2, N3, ...)
@Konrad Bańka - could be managed with CDS Day0/1 config - see below
Diagrams
Possible onboarding package content organisation
Non-RT-RIC Handled within ONAP thus R-APPS provided as DCAE mS
O-RAN sa ONAP Network Service composed of several xNF models
Each model could scale to different cloud regions creating new instances
Cross-References solving with CDS
Workflows
Static configuration that either xNF can't update on runtime or is rarely updated could be provisioned with Resource Assignment flow (For example initial replica numbers, interface versions or 5G Core Network endpoints). That config is (typically) gathered in the form of variables that are used during instantiation as Helm Variables (dedicated per vf-module)
Some configuration can be also processed just before (config assign) and just after (config deploy) xNF instantiation. This typically includes resolving payload and applying with Netconf/Ansible/Restconf calls to xNFs
Configuration Store - AAI (Runtime Instance (meta)data), C&PS (Configuration store)
@Konrad Bańka C&PS is (I think) supposed to allow cross-instance configuration view/edit and can be a good store for dynamic information such as related to dynamic number of xNF instances