This manual provides an overview of the TSUBAME3.0 system architecture, compute node configuration, software configuration, storage configuration, and how to obtain an account, log in, manage passwords, change the default login shell, use the storage system, software environment, job scheduler, and ISV applications.
This guide explains how to use TSUBAME3.0, a shared computing system at Tokyo Institute of Technology. It covers account setup, SSH login, storage systems (including Lustre and CIFS), software environments, and job scheduling with qsub and qrsh. Users can access compute nodes via SSH and use high-speed storage for simulations. The system supports ISV applications like ANSYS and MATLAB, as well as freeware tools for computational science.
The system architecture includes 540 SGI ICE XA nodes with Intel Xeon processors, NVIDIA GPUs, and 135 TiB of memory. Storage is managed via GPFS, Lustre, and local SSDs. Login nodes are restricted for non-interactive tasks to ensure fair resource use. Users must authenticate via SSH keys and can check their TSUBAME points and storage quotas using command-line tools.
| Compute Nodes | 540 SGI ICE XA nodes |
|---|---|
| CPU | Intel Xeon E5-2680 v4 (2.4 GHz, 14 core) x 2 per node |
| Memory | 256 GiB per node, 135 TiB total |
| GPU | NVIDIA TESLA P100 x 4 per node |
| Interconnect | Intel Omni-Path HFI 100Gbps |
| Operating System | SUSE Linux Enterprise Server 12 SP2 |
You need to upload an SSH public key and connect via SSH to one of the login nodes (login.t3.gsic.titech.ac.jp, login0.t3.gsic.titech.ac.jp, login1.t3.gsic.titech.ac.jp).
The system supports ISV applications like ANSYS, Fluent, ABAQUS, MATLAB, and freeware such as Gaussian, AMBER, and COMSOL.