Closed bnlawrence closed 4 years ago
Clearly some of them need to discriminate between the number recorded in the CMIP6 archive, and the number carried out. However, these are really not performance metrics, but rather experiment "support metrics". Need to think about how best to handle those.
As you have said (partially or wholly) elsewhere, Bryan ...
Pflops-hour, peak and linpack could be added to compute_pool (https://github.com/ES-DOC/esdoc-cim-v2-schema/blob/master/platform_classes.py#L39).
Model, CMIP6 config, platform are already in the CIM.
I'm not sure what to do about the others to do with aggregated amounts across all simulations in an experiment. Perhaps the Ensemble (https://github.com/ES-DOC/esdoc-cim-v2-schema/blob/master/activity_classes.py#L119) could have some "resources" properties added to it to capture them?
If we can get all of these into the CIM, compiling this entire table from the ES-DOC will be trivial.
I've added a new class in platform: ``project_cost'' which addresses some of these things, but not the linkpack, pflop stuff ... which belongs in machine. It's yuck. I'll add it there.
Might be an optional set of questions ... or a profile, or what. Anyway, these are they: