NTT docomo Business has launched the “GPU over APN Testbed,” an environment that enables the use of GPU resources distributed across the country as if they were a single platform, leveraging IOWN APN technology to achieve low-latency, high-speed, and high-capacity communication.
GPUs are distributed across eight locations: Sapporo, Kanazawa, Fukuoka, Osaka, and four sites in the Tokyo metropolitan area. The testbed allows for the validation of AI workloads—such as distributed AI training, distributed AI inference, and high-volume data transfers between sites using RDMA—by combining a GPU environment with 100 Gbps-class, low-latency, high-speed, and high-capacity communication.
Provided as a venue for joint trials to experience the utilization of distributed GPU resources, this initiative aims to advance the realization of the “AI-centric ICT platform” concept—a core component of the NTT Group’s AI-native infrastructure, “AIOWN.”

※Translating Japanese articles into English with AI