Parts/Operating systems/IBM Compute Node Kernel
Why this part
The Compute Node Kernel ran on the BlueGene compute nodes themselves while a conventional Linux layer handled I/O on separate nodes, splitting the machine into a part that runs your job with nothing else happening and a part that talks to storage and the network. Noise from an unrelated daemon waking up on a compute node is a real source of variance at BlueGene's scale, and CNK's minimalism exists specifically to remove it. BlueGene/L, Intrepid, Mira and Sequoia all ran it, and the design vanished with the product line: no equivalent shipped after Power BQC.
Systems using this part
- Systems
- 11
- Rmax underneath
- 31.48 PFlop/s
- Units recorded
- -
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Mira 2012-06 | Operating system | USA | - | 8.59 PFlop/s | IBM | #3 → |
| Sequoia 2012-06 | Operating system | USA | - | 17.17 PFlop/s | IBM | #1 → |
| Vulcan 2012 | Operating system | USA | - | 4.29 PFlop/s | IBM | #9 → |
| Intrepid 2008-01 | Operating system | USA | - | 458.61 TFlop/s | IBM | #3 → |
| Frontier2 2008 | Operating system | France | - | 95.45 TFlop/s | IBM | #14 → |
| JUGENE 2007-10 | Operating system | Germany | - | 190.90 TFlop/s | IBM | #2 → |
| Genius (Garching) 2007-10 | Operating system | Germany | - | 21.91 TFlop/s | IBM | #41 → |
| New York Blue 2007-06 | Operating system | USA | - | 82.16 TFlop/s | IBM | #10 → |
| BGW 2005 | Operating system | USA | - | 91.29 TFlop/s | IBM | #2 → |
| Intimidata 2004-12 | Operating system | USA | - | 13.78 TFlop/s | IBM | #42 → |
| BlueGene/L 2004-11 | Operating system | USA | - | 478.20 TFlop/s | IBM | #1 → |
“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.