Open Data
Everything on this site is rendered from static JSON files served from this domain. There is no API key, no rate limit, and no registration. Take the numbers and check them.
Data freshness unknown
Files
| File | Contents |
|---|---|
/data/index.json |
List of available month files |
/data/daily/YYYY-MM.json |
One record per day for that month |
/data/rollups/monthly.json |
Monthly aggregates |
/data/rollups/yearly.json |
Yearly aggregates |
/data/records.json |
All-time records and cumulative totals |
/data/inventory.json |
Live node and GPU inventory |
/data/growth.json |
New-user growth and retention cohorts |
/data/storage.json |
Filesystem capacity |
/data/pricing.json |
Cloud rate table used for cost avoidance |
/data/health.json |
Last collector run and per-source status |
A day record
{
"date": "2026-07-20",
"gpu_hours": {
"allocated": 1843.5, "available": 3072.0, "reported": 3360.0,
"down": 288.0, "idle": 1228.5
},
"cpu_hours_allocated": 7412.0,
"utilization": {
"available": 0.6001, "installed": 0.5486,
"availability": 0.9143, "from_sreport": true
},
"jobs": {
"total": 142,
"by_state": { "COMPLETED": 128, "FAILED": 9, "TIMEOUT": 5 },
"by_size": { "1": 84, "2-4": 41, "5-8": 17 },
"success_rate": 0.9014
},
"gpu_hours_by_model": { "h100": 921.0, "a100": 612.5, "l40s": 310.0 },
"gpu_hours_by_partition": { "general": 1712.0, "interactive": 131.5 },
"gpu_hours_by_qos": { "general": 1640.0, "interactive": 131.5, "protected": 72.0 },
"active_users": 37,
"wait_seconds": { "p50": 94.0, "p90": 1820.0, "p99": 14400.0, "samples": 142 },
"hourly_gpu_hours": [64.2, "... 24 entries ..."],
"records": {
"largest_job_gpus": 8, "largest_job_gpu_hours": 96.0,
"longest_job_hours": 23.8, "max_nodes_in_job": 2
},
"groups": [
{
"name": "Example Lab", "department": "Statistics",
"division": "Physical Sciences Division", "type": "lab",
"gpu_hours": 421.0, "cpu_hours": 1684.0, "jobs": 22, "users": 5
}
]
}
What you will not find
No usernames, job IDs, job names, exit codes, or node names. No per-person
series of any kind. A research group appears by name only if an operator
added it to the allowlist and it clears the anonymity threshold for the period
shown; everything else is aggregated into "Other". The
methodology explains the guarantees and
how they are enforced.
Reuse
These are public University figures — use them in proposals, reports, and talks. A link back to this site is appreciated so readers can find the methodology. Charts on this site each have a download button in their top right corner that exports a PNG at 2× resolution for slides.
Source code and data history: github.com/dsi-clinic/cluster-impact.