随着AI大模型的快速发展，算力需求急剧增加，聚焦车企普遍关注的先传后训方式导致算力资源利用率低、敏感 数据出园区导致安全担忧等痛点，提出了一种基于存储拉远和无损网络技术的解决方案，通过构建智算实验网，采用边传边训 的创新模式，实现了数据实时处理与模型快速更新。实验结果表明，该方案显著提升了算力使用效率，单任务训练时间缩短 50%，数据传输速率提高至7.3 Gbps，且在200 KM拉远条件下仍能保持高效运行。然而，未来仍需关注数据安全与隐私保护等 挑战。本研究为应对大模型时代的算力需求提供了新的思路和方法。

With the rapid development of AI large models, the demand for computing power has sharply increased. Focusing on the pain points of low utilization of computing power resources and security concerns caused by sensitive data leaving the park, this ar‐ ticle proposes a solution based on storage and lossless network technology. By constructing an intelligent computing experimental network and adopting an innovative mode of simultaneous transmission and training, real-time data processing and rapid model up‐ dates have been achieved. The experimental results show that this scheme significantly improves the efficiency of computing power usage, reduces single task training time by 50%, increases data transmission rate to 7.3 Gbps, and can still maintain efficient opera‐ tion under 200 KM distance conditions. However, challenges such as data security and privacy protection still need to be addressed in the future. This study provides new ideas and methods for addressing the computing power demand in the era of large models.