縮退運転フェーズ/Reduced Operation Phase
KAGAYAKI-CPUが97ノードでの縮退運転を開始しました。
KAGAYAKI-CPU has resumed operation with 97 nodes.
混雑が予想されますが、皆様のご協力をよろしくお願いいたします。
It is expected to be crowded, and we appreciate your cooperation.
Service end for system replacement of HPC System "KAGAYAKI"
HPC System "KAGAYAKI" will be replaced with a new system at the end of next February, as the other computing facilities had been done so. Since the replacement will include related facility work such as updating the cooling system, the current KAGAYAKI is expected to be taken out of service at an early date.
We will update this page as soon as the details are finalized, so please check back regularly.
In addition, please carefully consider your schedule when using the KAGAYAKI for papers, research conferences, etc.
If you are using the KAGAYAKI jointly with researchers outside the university, please share this information with your joint research partners as necessary.
Upcoming HPC and AI System "HAKUSAN" (2026/03/02 Update)
■ Name of the New System
HPC and AI System "HAKUSAN"
Future announcements and documents will refer to the system using this name (HAKUSAN).
■ Job Scheduler
The job scheduler will transition from the current PBS to Slurm.
We plan to provide documentation and hold workshop on how to use Slurm and write job scripts around the time of service launch.
■ Service Start Schedule
Due to exterior wall construction on Information Building 2, the installation process is currently running approximately one month behind the original schedule.
We are targeting a service launch within March, but the specific start date has not yet been determined.
We will inform you of the official service start date via MPC Teams as soon as it is confirmed.
■About MPC Teams
JAIST's large-scale computing clusters are generally available to anyone with a JAIST account without requiring separate application.
However, to ensure smooth operations, we ask all users to join MPC Teams for announcements and information exchange among users.
We will also use MPC Teams for announcements regarding the launch of this system.
We welcome anyone interested in HAKUSAN to join.
MPC Teams
We will continue making preparations toward the service launch, and we appreciate your continued support.
Service Suspension and Reduction Due to KAGAYAKI Replacement (2025/11/14 Update)
Overview
To build the new system, dismantling of the current system will begin in mid-December.
However, since a prolonged suspension of computing services is expected to have a significant impact on research activities, we plan to relocate part of KAGAYAKI and operate it in a reduced configuration.
Request to Users
During reduced operation, the number of CPU nodes will be about one-third of normal, and congestion is expected.To ensure priority for time-sensitive research, we ask those with flexible schedules to postpone their usage during this period.Additionally, please note that there will be a period when KAGAYAKI is completely shut down for preparation.During this time, running jobs will be forcibly terminated, and GPU nodes and VM hosting services will also be unavailable.We apologize for any inconvenience this may cause.We are making every effort to minimize downtime and ensure a smooth migration.Thank you for your understanding and continued cooperation.
Schedule (as of November 14, 2025)
〇Normal Operation Phase (until November 27, 2025)
KAGAYAKI will be available as usual with 280 nodes.
- KAGAYAKI-CPU: Normal operation
- KAGAYAKI-GPU: Normal operation
- Lustre Storage: Normal operation
- VM hosting services: Normal operation
〇Partial Reduced Operation Phase (November 28 – December 16, 17:00)
To shorten the complete shutdown period, some nodes will be removed in advance for reduced operation.
- KAGAYAKI-CPU: Reduced operation using 174 nodes
- KAGAYAKI-GPU: Normal operation
- Lustre Storage: Normal operation
- VM hosting services: Normal operation
〇Full Service Suspension Phase (December 16, 17:00 – December 22)
Management nodes and Lustre storage will be relocated in preparation for reduced operation.
Jobs running at the start of this phase will be forcibly terminated.
- KAGAYAKI-CPU: Service suspension
- KAGAYAKI-GPU: Service suspension
- Lustre Storage: Service suspension
- VM hosting services: Service suspension
〇Reduced Operation Phase (December 22 – around the end of February 2026)
Limited operation will be provided with 97 CPU nodes and GPU nodes.
Since the number of nodes will be about one-third of normal operation, congestion is expected.
* Available resources may increase or decrease depending on conditions such as cooling capacity.
- KAGAYAKI-CPU: Reduced operation using 97 nodes
- KAGAYAKI-GPU: Normal operation
- Lustre Storage: Normal operation
- VM hosting services: Normal operation
〇New System Phase (from March 2026 onward)
The new computing system is scheduled to begin service.
Details will be announced as soon as they are finalized.
*The following is a guideline as of August 2025.
This web page will be updated as soon as details are finalized.
Scheduled to end of KAGAYAKI
Around mid-December 2025 to January 2026
Scheduled to start next HPC system
Around March 2026
Subject to renewal
- CPU Node 280 nodes
- Network File Service for Research (Lustre)
The following servers are not subject to this renewal, but may temporarily stop due to the update of the Network File Service for Research (Lustre).
- GPU Node 1
- GPU Node 2
- VM Large Memory Node
- VM GPU Node-S/L
- VM General Purpose Compute Node
About data migration for "Lustre" storage
Details will be updated as soon as they are finalized.