Parts/Storage/WEKA Data Platform
Why this part
Software rather than an appliance, designed around the specific failure mode of GPU training: thousands of tiny random reads that a filesystem tuned for large sequential HPC writes handles badly. A GPU that is waiting for data is the most expensive idle asset in a data centre, and the entire commercial case for this category is that storage is now worth optimising because the thing it feeds costs so much.
Systems using this part
- Systems
- 7
- Rmax underneath
- 79.46 PFlop/s
- Units recorded
- -
0 of 7 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Sunrise 2026-06 | Storage | United Kingdom | - | 34.77 PFlop/s | WEKA | - |
| MUSICA 2024 | Storage | Austria | - | 31.84 PFlop/s | WEKA | #50 → |
| Mahuika 2018 | Storage | New Zealand | - | - | WEKA | - |
| Redtail | Storage | USA | - | 12.85 PFlop/s | WEKA | - |
| Skipjack (JHU) | Storage | USA | - | - | WEKA | - |
| Coral | Storage | USA | - | - | WEKA | - |
| SLAC S3DF | Storage | USA | - | - | WEKA | - |
“Units recorded” sums only the edges where a public source states a quantity. It is a floor, not a total, and should never be read as installed base.