P1 Compute
Requests for comments
Compute is central to research. RFCs are how the research community and providers agree on what researchers need from it.
Open RFCs
- RFC 0002 A helper tool for every researcher Why P1 offers a small, optional tool for interactive GPU sessions, so every researcher gets the same safe way in. Open
- RFC 0001 Data classification How researchers classify data that is not public, and which resource, controls and people each class allows. Open
See all RFCs, with their status, or follow changes in the Atom feed.
How RFCs work
| Who | Does |
|---|---|
| Researchers | Say what they need: in the compute needs survey, at events, in the P1 consortium. Nobody has to write an RFC. |
| Research engineers | Draft RFCs from that feedback, with an open network of DevOps, MLOps and domain experts. |
| HPC governance board | Decides each RFC. In the P1 consortium it has one PI per partner institution, and they carry the mandate regarding compute decisions. |
| Providers | Review drafts and meet the requirements. |
Each RFC covers one concern.
Why RFCs
- Listening: the needs of researchers set the direction.
- Expectations: providers know what access and service the research community expects.
- Open knowledge: no standard depends on a single consultant or vendor.
- Both ways: the research community trains researchers to use shared compute well through teaching, events and innovation.
- Best effort: nothing here is enforced. It works because everything is public.