Core Viewpoint - Lenovo's innovative RNL technology addresses long-standing challenges in load balancing for AI training and inference scenarios, particularly in RoCE networks, as recognized by its acceptance at the IEEE CyberSciTech 2025 conference [1][2]. Group 1: Technology Innovation - The RNL technology incorporates a closed-loop system of "multi-dimensional perception + path load balancing + incremental migration," combining algorithmic innovation with practical value [2]. - The multi-dimensional perception mechanism allows real-time awareness of network topology, AI task network demands, and RoCE link load status, providing a data foundation for dynamic scheduling [2]. - Path load balancing optimization utilizes virtual-physical network mapping and path scoring algorithms to intelligently select optimal data transmission paths, maximizing bandwidth utilization [2]. - Incremental flow migration employs a strategy to avoid instantaneous delays during link traffic adjustments, ensuring business continuity [2]. Group 2: Future Plans - Lenovo plans to extend the RNL technology to high-performance storage and HPC scenarios, incorporating deep learning algorithms to enhance congestion prediction capabilities [2]. - The company aims to validate the comprehensive performance of RNL technology in large AI clusters with thousands to tens of thousands of nodes, continuously driving innovation and iteration in AI network technology [2].