Case Study: Tokyo Tech achieves high-performance, energy-efficient AI and big-data research (15.9PB, 150GB/s) with DataDirect Networks

Changing Research With a Forward-Looking AI and Big Data Computing Infrastructure

Tokyo Institute of Technology (Tokyo Tech), Japan’s largest science and technology university, built the innovative TSUBAME3.0 supercomputer to support AI, machine learning and big‑data research. The challenge was to dramatically speed parallel data access and keep improving algorithm interaction with storage while minimizing power consumption and maximizing system efficiency.

DDN supplied a 15.9 PB Lustre parallel file system (three ES14K appliances) with ExAScaler and IME Flash‑Native Data Cache, delivering a theoretical peak throughput of 150 GB/s. The deployment accelerated data access for data‑centric and AI workflows, established a roadmap for next‑generation research computing, and opened new application areas such as graph computing and machine learning.

Tokyo Tech