Anonymized HPC Job Records from Kestrel System

Anonymized job-level records from the Kestrel high-performance computing system, covering jobs submitted between August 2023 and December 2025. Each record represents a Slurm batch job and includes scheduling metadata, resource requests, utilization, energy consumption estimates, and computed efficiency metrics.

Use Cases

• Analyze resource utilization patterns against resource requests to identify inefficiencies in HPC job scheduling.
• Model energy consumption estimates based on job scheduling metadata and resource usage to forecast system power demands.
• Compute efficiency metrics for different job categories to optimize HPC cluster performance and resource allocation.
• Study anonymized job submission patterns over time using scheduling metadata to understand workload seasonality.

Strengths

• Data covers a substantial timeframe of over two years, from August 2023 to December 2025.
• Includes multiple analytical dimensions: scheduling metadata, resource requests, utilization, energy estimates, and efficiency metrics.
• Personally identifiable fields have been anonymized via cryptographic hashing to protect privacy while preserving data utility.

Limitations

• Sample size and specific row count are unknown, limiting assessment of statistical power.
• Finer-grained job_step records containing per-step resource usage are not included in this data package.
• Data is specific to the Kestrel HPC system at the National Laboratory of the Rockies, which may limit generalizability.

Provenance

Source

National Laboratory of the Rockies (NLR) HPC Operations and Data Analytics teams.

Collection Method

Derived from the Kestrel schema of the NLR HPC job database, with sensitive fields hashed for anonymization.

Time Range

2023-08 to 2025-12 (Mountain Time zone).

Freshness

Data was last updated in March 2026 and covers events up to December 2025.

Geography

null