High Performance ComputingResearch

National Research Institute HPC Facility

Maximizing Scientific Discovery per Kilowatt-Hour

ResearchHPCScientific ComputingClimate ModelingGenomics

Scientific Output

+67%

Increased compute jobs completed

Queue Time

-71%

Reduced from 8 weeks to 16 days

Energy per FLOP

-28%

Improved computational efficiency

Research Impact

+45%

Publications citing facility resources

The Challenge

A national research institute operated a petascale computing facility supporting climate modeling, genomics research, and physics simulations. Fixed energy budgets limited compute availability, creating 8-week queues for large jobs. The facility needed to maximize scientific output within congressional budget constraints.

The Solution

Helios implemented science-aware resource optimization that prioritized workloads based on research impact and energy efficiency. The platform deployed intelligent scheduling that matched job thermal profiles with available cooling capacity. Energy-aware allocation enabled fair distribution across research programs while maximizing total scientific output.

Implementation Timeline

Research Assessment
4 weeks
Understanding scientific computing patterns
Scheduler Integration
6 weeks
Connecting to existing job management
Allocation Framework
4 weeks
Implementing fair-share energy model
Researcher Enablement
3 weeks
Training scientists on new capabilities

Key Insights

Science-aware scheduling prioritized high-impact research during peak efficiency periods
Thermal-aware job placement eliminated 96% of cooling-related throttling
Fair allocation model resolved inter-program resource conflicts
Energy transparency enabled researchers to optimize their own workflows

Measurable Impact

Quantified results from this transformation