Adaptive xApp Architecture for Efficient E2 Interface Management and Near-Time Data Streaming in O-RAN

There is a lack of adaptive mechanisms for handling dynamic E2 subscriptions in O-RAN architectures to enable high-volume message processing in near-real-time RIC environments. In this paper, we present the design and implementation of a high-performance xApp for near-real-time RAN control within the O-RAN architecture. The xApp manages E2 subscriptions dynamically, processes incoming KPM indication messages in a multi-threaded manner, and forwards enriched metrics to message brokers for cloud analytics. It employs closure-based callbacks to inject E2 node identifiers into messages without modifying the standard interface, unlike prior studies that break standardization efforts. The system is implemented in C using the libffcall library to enable closure functionality. The xApp is evaluated in a Kubernetes environment with FlexRIC and multiple simulated gNBs. Evaluation shows that the proposed architecture achieves significant performance gains over a single-threaded baseline. The average end-to-end latency decreases from 47 ms to 14 ms, and throughput increases from 820 to 3150 messages per second. These results demonstrate that the proposed design provides lower-latency data handling for near-real-time RIC applications.