Accelerating Inter-App Communications for Time-Critical O-RAN Control Loops

The Open Radio Access Network (O-RAN) is reshaping cellular architectures through disaggregation, openness, and programmability, enabling intelligent control across modular RAN components. A growing ecosystem of user-defined control-plane applications, namely rApps, xApps, and the emerging dApps, operates at different timescales and unlocks advanced control loop capabilities, but introduces diverse and stringent Quality of Service (QoS) requirements for communication. Current implementations typically rely on cloud-based messaging systems, which privilege transparency and ease of use for developers and limit support for time-sensitive workloads. In this paper, we propose that O-RAN control loops base their internal communication on INSANE, a cloud-native, data-centric middleware that supports multiple networking stacks, including kernel-bypass option such as eBPF XDP, DPDK, and RDMA. Our preliminary evaluation shows that INSANE achieves nearly 2× higher throughput than widely used alternatives, while reducing 99.9th-percentile latency of over an order of magnitude for small messages. At the same time, INSANE preserves a uniform and easy-to-use programming interface. These results highlight INSANE as a promising foundation for faster, more predictable control loops, a significant step toward the ultimate goal of enabling AI-driven RAN optimizations in O-RAN systems.