GPU Computing Infrastructure 1 Set

2026-11-03

Classification of the products to be produced : 1 Nature and quality of the products : GPU Computing Infrastructure Set Type of the procurement : Purchase Basi requirements for the procurement : The following basi requirements must be met A The procured system shall comprise a shared GPU computing platform for research, education, secure data analysis, and related purposes. It shall consist of GPU-equipped compute nodes, non-GPU compute nodes, storage management nodes, and large-capacity storage, and shall be capable of stable and integrated operation as a unified system. B The compute nodes shall be primarily based on CPUs such as AMD EPYC processors and provide an aggregate computing capacity of approximately 20 CPU cores or more. GPU-equipped nodes shall include approximately to 1 data center-class GPUs and support workloads such as dee learning, machine learning, training and inference of large language models (LLMs), and data analytics. C The system shall provide at least TB of main memory, with expandability to approximately TB as required. In addition, node-local storage, including both compute and system partitions, shall be configured for high-performance access and may employ NVMe SSDs or HDDs depending on usage requirements. D The storage infrastructure shall be capable of reliably storing both current and future large-scale datasets. It shall accommodate at least several hundred terabytes of existing data and university-wide shared data and provide scalability to an effective capacity of approximately PB. E GPU servers and storage nodes shall be interconnected through a high-speed network of at least 1 Gbps. Furthermore, sufficient bandwidth and network redundancy shall be ensured for connectivity with the university network. F The system shall support the use of software such as Slurm for job scheduling and management, Apptainer for containerized execution environments, and Open OnDemand as a user-facing operational interface. In addition, it shall provide functions that support secure operations, including the clearing of GPU memory and related resources after job completion. G The system shall enable the secure execution of research activities, educational use, and the analysis of sensitive data through user-based access control, separation of usage domains, log management, and VPN connectivity where required. H The system shall provide scalability for future GPU expansion, storage expansion, node additions, and configuration changes, and shall support phased implementation and incremental system growth. Time limit for the submission of the requested material : 1 : 00, October, 202 Contact point for the notice : Takashi Fujita, Financial Management section, Niigata University, 805 Ikarashi 2-no-cho Nishi-ku Niigata-shi 950-218 Japan, TEL 025-262-7671

Japan

2026-09-01