OCUDU Ecosystem Developer Summit

As the OCUDU ecosystem matures, validating Open RAN performance requires shifting from isolated component testing to verifying end-to-end interoperability. The current Retina testing framework is limited by tight coupling to internal processes, restricting its ability to orchestrate external components like diverse 5G Core implementations and O-RU hardware.
This presentation, a collaborative effort between the UNH Interoperability Labs (UNH-IOL) and Wind River, proposes a strategic roadmap to evolve the Retina framework. We outline an architectural shift toward a modular, "building-block" approach where deployment of RAN components, Core networks, and physical radio units is decoupled from the test automation harness. By leveraging Ansible-driven orchestration to manage diverse deployment stacks, this extension enables vendor-neutral integration testing and allows common test scenarios to be shared between component and end-to-end testing. The session will review completed framework changes and our road map, while seeking community feedback to ensure the framework supports diverse infrastructures and open-source projects.

As OCUDU matures into a carrier-grade open-source 5G stack, operators and integrators need vendor-neutral, reproducible evidence of how O-RU hardware behaves in a real OCUDU deployment—not just vendor-reported claims.

ITRI Lab, an OCUDU Lab partner in the APAC region, is building a neutral validation platform that benchmarks multiple RU architectures on OCUDU, using a methodology covering Open Fronthaul interoperability, throughput/latency KPIs, multi-UE scenarios, and long-duration testing—referenced where applicable against TIP baseline configurations.

This session covers the platform and methodology, selected multi-vendor observations, and a deep-dive validation case: an 8T8R Category B O-RU from LIONS Technology integrated with OCUDU, including Open Fronthaul behavior and beamforming performance. This APAC lab also validates contributions from Taiwan's broader Open RAN ecosystem—LIONS extending into CBRS-compliant PAL/GAA deployments, and G-Reigns on CU/DU integration—into OCUDU, with further cases to follow.

Attendees leave with a reusable framework for OCUDU RU validation, a reference against common baselines, and concrete multi-vendor benchmark results on OCUDU.

Problem: Open-source O-RAN software wants to become carrier-grade, but there's no agreed way to prove it got there. Coverage gets claimed, not verified. Scanners can under-rate severity. Code that works in simulation can fail against real hardware.

Solution: We play two roles: independent evaluator and remediation partner. We test and report what we find, then help fix it with the builders, and retest before calling anything resolved.

How It Works: Every test traces to a spec requirement. Coverage is tagged honestly, verified vs. not observable, not rounded up to "tested." Scanner severity is rechecked and escalated, never downgraded. Testing runs on real hardware, not simulation. Fixes get rerun and reverified.

Current Status: Over a hundred executions since early this year across functional, security, and fronthaul testing. Coverage accountability moved from two-thirds to nearly all of it. Our latest security run found real failures, published as-is, remediation underway.

Next Steps: This test bed isn't tied to one program. Once these gaps close, it extends to other open O-RAN components, deeper fronthaul coverage, and eventually 6G, a model the community can reuse.

A commercial O-RU leaves the factory unreachable, and the step that carries it to a first management session is not published. The OCUDU ecosystem already ships lab-grade pieces of that step, a Kea DHCPv4 service encoding Option 43 and an O1 adapter that accepts SSH call-home. Dorado will present what commercial multi-vendor hardware still takes: DHCPv6 with Option 17 for v6-only radios, CMPv2 enrollment with TLS call-home on top, one vendor profile per device, and G.8275.1 timing inside the chain.

Results cover connecting multiple commercial O-RUs to a complete end-to-end OCUDU in a captive environment. A minimum of three commercial O-RUs reach stable call-home with no human step after cabling, with executable O-RAN M-Plane conformance and security cases in Robot Framework as the CT stage. Work is ongoing and additional results may be discussed.

OCUDU is emerging as a key open-source foundation for carrier-grade, research-oriented 5G. While most demos remain lab-based, real-world field trials are critical for understanding operational and interoperability challenges.
We present SUTD's OCUDU deployment journey, from multi-vendor O-RU interoperability testing (e.g., Sera Network, Metanoia-based, Compal, QCT, Pegatron) to a fully operational outdoor Open RAN system supporting 5G-connected drone communications.
Next, we share performance results from an outdoor Pegatron O-RU deployment at SUTD Drone Arena, using real-time drone connectivity tests to evaluate mobility, coverage, throughput, and robustness outdoors—a realistic assessment beyond controlled labs.
Finally, we demonstrate OCUDU's integration with the NeuroRAN's SMO/Non-RT RIC and Near-RT RIC via standardized O1/E2 interfaces. This enables AI-driven closed-loop optimization apps, including an Energy-Saving rApp and end-to-end Network-Slicing solution including Slice Resource Management-rApp and Network Slicing xApp.
We close with lessons learned, interoperability challenges, OCUDU roadmap recommendations, and opportunities to accelerate AI-native Open RAN adoption.

Cohere’s Universal Spectrum Multiplier (USM) product is designed as an adjunct to a gNB that performs channel analysis and codebook-free MU-MIMO scheduling based on continuous calibration of the transmission chain and measuring the geometric reciprocity of the channel itself. The USM connects to gNB via gRPC wrapping the E2SM-LLC API. We’ll start the session with an introduction to how the architectural building blocks of the USM, gRPC/API, and gNB function together to add spectral capacity. We will also be talking about how Cohere enabled MU-MIMO support in OCUDU stack.

OCUDU gives the Telecom industry a 3GPP and O-RAN compliant, open-source CU/DU stack with multiple Internal Splits and Open Fronthaul Split options. The OCUDU stack is built over a very capable Software architecture, which is easy to scale. But the user plane - specifically the Crypto engine and the N3/F1-U transport interfaces, still rely on general-purpose software primitives that are designed for limited throughput. The mbedTLS based Crypto processes one packet at a time with no session persistence and no batch processing. The N3 and F1-U use kernel UDP sockets with multiple memory copies, syscall overhead, and I/O interrupts. This results in throughput capping at 2–5 Gbps per core and limits the user-plane scalability.
Here is a 2-phase proposal to replace these bottlenecks with DPDK-based alternatives, while preserving existing implementations as fallback option.
Phase 1 - DPDK Crypto Engine. Replace the mbedTLS crypto path in PDCP with DPDK Cryptodev. Today, OCUDU’s security_engine_impl performs:
Phase 2 - DPDK-Based N3 & F1-U Transport (Evolution). Replace the kernel UDP socket interfaces for N3 (toward UPF) and F1-U (toward O-DU and inside O-DU) with DPDK poll-mode drivers.

In this technical and operational case study, the Marine Corps Tactical Systems Support Activity (MCTSSA) and Naval Information Warfare Center (NIWC) Pacific Tactical Cellular Lab (TCL) will present how they deployed OCUDU to provide 5G at the tactical edge for expeditionary military operations. During a military exercise in the spring of 2026, the TCL rapidly installed and deployed a gNB on a military base in the Western Pacific using commercial-of-the-shelf equipment. The deployment of OCUDU, Open5GS, and an O-RU (n77) from Dell Technologies provided mobile access to the Marine Corps' tactical network across an entire airfield. Topics of interest in the discussion will include: (1) Distributed core network architecture and network slicing, (2) Benefits of ORAN/OCUDU for new RU integration, (3) Hardware considerations to support expeditionary operations, (4) Future tactical use-cases for OCUDU and ORAN, (5) Importance of interoperability with O-RAN standards and (6) Challenges and areas for improvement to use OCUDU to provide 5G coverage at the tactical edge.

Integrating OCUDU for research purpose is quite simple, however a mission critical deployable system, especially in constrained environments introduces a complex set of engineering challenges. At SEMPRE, we have taken OCUDU from source code to deployed edge systems supporting commercial and mission-critical defense applications. This session will discuss hardening OCUDU for constrained edge environments, including software packaging and configuration, zero-trust architectures, timing and synchronization in GNSS-denied conditions, and disconnected operations, field performance, fault recovery, lifecycle management, and AI-assisted tools for unattended operations.
We will also share practical lessons from system bring-up, platform integration, performance tuning, and the stability challenges encountered moving OCUDU into operational environments. We will highlight how the same hardened baseline can support capabilities beyond connectivity, including ISAC-based small UAS detection and multi-sensor fusion. Finally, we will identify opportunities to reduce deployment complexity, improve production readiness, and make OCUDU easier to adopt across for enterprise and defense use cases.

Operating secure and resilient open-source RAN stacks such as OCUDU remains challenging, particularly for fine-grained runtime observability as well as real-time detection/mitigation of faults and security threats. This presentation introduces an AI-driven Security Operator (AISecOps) framework designed to harden the OCUDU architecture with self-healing and self-defending capabilities.

At its core the AISecOps operator leverages extended Berkeley Packet Filters (eBPF) technology to non-invasively extract real-time RAN-to-Core control plane and KPM telemetry, requiring zero source code modifications. This telemetry feeds a neurosymbolic diagnostic engine that combines deep-learning-based anomaly detection with reasoning-based safety guardrails to detect runtime fault and threat events. Moving beyond passive alerts, a knowledge-grounded agentic AI operator delivers autonomous threat incidents analysis, root-cause diagnostics, and mitigation response.

Attendees will learn how to deploy eBPF into OCUDU for deep security observability, and how agentic AI can automate the security incident handling cycle and execute fully autonomous or human-in-the-loop remediation at the edge.

For many OCUDU target deployments an outage is a safety event, not just a bad customer experience: factory floors here robots and machine-vision safety loops depend on the network for low-latency control, and military and tactical sites running drone and robot command-and-control and ISAC. End users in those settings stop asking only about throughput and coverage and start asking two harder questions: can you prove it is secure, and can you keep it running while someone is attacking it?
This talk treats security and availability as one problem and proposes a common, modular security-validation framework that any OCUDU-based gNB product can adopt across vendors. I will walk the full trust chain (operator → SMO → management plane → gNB → open fronthaul → O-RU → edge), cover DoS resilience, provide a gap analysis against FIPS, DISA STIG and Common Criteria, and share a prioritized roadmap for hardening 5G and beyond Radio Access Networks built on OCUDU.
Attendees will leave with a proposed security hardening areas, a concrete understanding of where FIPS/STIG/Common Criteria apply, and a starting definition of a shared validation framework they can adopt and contribute to.