In this report, we evaluate the feasibility of remote storage for AI training by connecting two sites via APN and using image data for Unet-3D model training to assess storage access performance. Through this, we aim to verify the effectiveness of APN in maintaining practical training time in a distributed environment.