Changes to the ASPIRE 2A+ System from 2 Jun 2025
Dear ASPIRE2A+ Users,
We would like to inform you of upcoming changes to the ASPIRE 2A+ system, effective 2 June 2025. These updates are part of our ongoing efforts to improve resource management, performance monitoring, and user experience.
What’s Changing?
1.PBS Queue Structure Update
Starting 2 June 2025, all jobs submitted to the normal queue will be automatically routed to one of the following new queues based on the job’s characteristics:
- aidev Queue – For development and interactive workloads
- Max Walltime: 12 hours
- GPU Limit: Up to 8 GPUs
- Ideal for: Testing, debugging, and short interactive sessions.
- aiq1 Queue – For small batch jobs
- Max Walltime: 24 hours
- GPU Limit: Up to 7 GPUs
- Ideal for: Short production runs and low-GPU batch jobs.
- aiq2 Queue – For large batch jobs
- Max Walltime: 120 hours (5 days)
- GPU Requirement: Minimum of 8 GPUs
- Ideal for: Long-running, resource-intensive jobs.
What You Need to Do:
Continue submitting jobs to the normal queue as usual. The system will automatically route your job to the appropriate queue based on the requested GPU count and walltime. We encourage reviewing your job scripts to ensure that they align with the new queue definitions.
2.Change in dcgmi Access Permissions
From 2 June 2025, the dcgmi (NVIDIA Data Center GPU Manager) tool will no longer be executable by regular users.
This change is necessary to ensure the accuracy of system-wide, system-level GPU performance monitoring. If your workflows rely on dcgmi, please contact the our Helpdesk for alternative solutions or assistance.
Summary of Actions
- Review and adjust your job scripts to align with the new queue definitions.
- Avoid using dcgmi in user applications or scripts.
- Reach out to the Helpdesk if you need help adapting to these changes.
Thank you.
Warm regards,
The NSCC Team