How to represent which compute provider was used?

[mr-c]

Could be for the entire workflow and/or each individual step (for the case of distributed execution)

• Where (and with what term) in the ro-crate-metadata.json metadata file would we represent this for an entire workflow?
• Where (and with what term) in the ro-crate-metadata.json metadata file would we represent this for a specific step?
• How should the provider be credited? Current thinking is to use an ROR + a local identifier (like a cluster name)
• Probably should be a list of identifiers, for layered systems. Example: OpenStack VMs provisioned on the de.NBI cloud, at the Freiburg University instance. (needing to give credit to OpenStack, de.NBI Cloud, and the Freiburg Uni computers)

(@mr-c has heard about instrument identifiers, but knows nothing about that beyond the existence of the concept)

[mr-c]

I see the instrument field of the Process Run Crate profile, but that seems to be capturing the name of the tool, not where it was run

https://www.researchobject.org/workflow-run-crate/profiles/process_run_crate/#requirements

[jmfernandez]

I guess agent is most suitable for that task. When it is used, it should link to the Person or the Organization.

The question here is whether an Organization entity is suitable for the purpose.

[stain]

From https://www.researchobject.org/workflow-run-crate/profiles/provenance_run_crate/ figure
I think it would be better to attach to the ControlAction as there may be some kind of job submission, containers etc. involved before it comes down to the file-based command line execution represented by the bottom CreateAction for the tool. Perhaps the new https://schema.org/provider can be better used -- it currently points to a Person or Organization so same issue as with agent. (agent I would say would rather be the workflow engine in that case, but it's not as needed information as the reverse link with object is already there.)

[stain]

If you want to identify the infrastructure separate from the infrastructure organisation, (e.g. ARCHER2) then perhaps we need to find a way to a https://schema.org/Service or even https://schema.org/WebAPI (e.g. WES)

[stain]

location on the tool execution could also be used, perhaps to a https://schema.org/VirtualLocation ?

[stain]

Discussion in ELIXIR Compute WP5, use case is Mgnify workflow in Galaxy, executed in Pulsar nodes. From there we would ideally need:

• Where was it run? (identifier/name of the platform)
  • Galaxy (version NN), at usegalaxy.eu; using pulsar nodes XX & YY; final storage at de.NBI Cloud host ZZ
  • Workflow system/platform & version
  • Host name & organization identifier (ROR or something for consortia/virtual orgs)
  • Compute providers (if different from above) (Pulsar nodes)
  • Storage providers (if different from above)
• Which workflow? (identifier/name like from workflowhub.org)
• Which user-supplied data was used?
• ENA Sample accession identifiers, all data has to be in ENA first. Example: https://www.ebi.ac.uk/ena/browser/view/SAMN11835499
• What user-chosen parameters?
  galaxy/CWL input object (key-value pairs)

Raised by @mberacochea