Skip to content

MVP Scope And Roadmap

MVP Now

  • one node or tightly scoped lab deployment
  • one DU-high runtime
  • one cell group
  • one UE attach plus ping path
  • canonical IR and backend profile selection
  • stub_fapi_profile, local_fapi_profile, and aerial_fapi_profile contract paths
  • ranctl command contract and approval model
  • observability and artifact capture skeleton

Next

  • push the local DU-low Port path from synthetic sidecar toward a real native worker
  • add config validation and release packaging
  • add integration tests for drain, switch, verify, and rollback
  • model backend health and degraded states more concretely
  • deepen the cumac_scheduler contract host toward a real external scheduler worker
  • keep the core AMF SCTP edge bounded and explicit while it moves from flat JSON intent to a bounded NGAP-shaped JSON envelope for NGSetup, InitialUEMessage, UplinkNASTransport, DownlinkNASTransport, PDUSessionResourceSetup, and UEContextRelease
  • keep the core SMF and UPF-control PFCP edge bounded and explicit while it covers the implemented Create*/Modification subset and bounded Remove* grouped-IE handling before any broader PFCP parity claim
  • keep subprojects/ran_replacement/ as a separate design-first track for an OAI CU/DU replacement targeting one n79 real-RU and real-UE attach-plus-ping lane against a real Open5GS core, with an agent-friendly ranctl control surface before runtime cutover

Later

  • multi-cell and multi-DU orchestration
  • handover support
  • advanced scheduling coordination
  • real DU-low implementation
  • real NVIDIA Aerial adapter
  • real cuMAC scheduler adapter
  • live SCTP, NGAP, F1AP, E1AP, and GTP-U integration

Out Of Scope For This Bootstrap

  • production timing guarantees
  • PHY or low-PHY implementation
  • vendor-specific Aerial internals
  • distributed control plane clustering decisions

Design-first Open RAN architecture documentation.