Add system architecture diagram to README

[dlq]

Motivation

• Improve documentation by adding a high-level system architecture diagram to make component roles and interactions clearer.
• Provide a visual overview that shows how BrainPortal, NeuroHub, Bourreau, data providers, and HPC resources relate.
• Help new contributors and operators understand the overall deployment and data/job flow.

Description

• Added a new ## System architecture section with a Mermaid flowchart diagram to README.md that models users, NeuroHub, BrainPortal, the shared database, data providers, Bourreau, HPC scheduler, compute nodes, and shared scratch.
• Included an explanatory paragraph describing how BrainPortal orchestrates data access and delegates execution to Bourreau, and how Bourreau interacts with HPC schedulers and shared storage.
• The change was applied to README.md and committed with the message Add system architecture diagram to README.

Testing

• No automated tests were run for this documentation-only change.
• Repository CI (badge present in README) will exercise automated checks on the PR when submitted.

---

Codex Task

[MontrealSergiy]

I think it is a great addition, but media (mermaid) is not advanced enough for such a complex software.

A safer approach I think is cut and paste structural diagram(s) verbatim from Pierre presentations, conference poster or other publications.

As it was noted in slack, the accuracy is questionable.

• I think Datasets are accessed via data Providers,

• NeuroHub and Cbrain UIs are almost parallel, both CBRAIN Portal and NeuroHub Portal are rails apps, which however share same db and resources.

[MontrealSergiy]

verbiage needs a check

e.g, in

Bourreau can also fetch and stage data from providers as part of backend task execution.

can is redundant if not misleading

[MontrealSergiy]

ok with me

[MontrealSergiy]

normally bourraux do not use ssh to access Slurm, please correct

'

[MontrealSergiy]

Please assign yourself, add Documentation label
to PR

[MontrealSergiy]

Just FYI, there could be local data providers and Bourreau ( do not add them to the diagram, it is already big)

I would have two separate External Compute and External Storage resources (flat is better than nested)

[MontrealSergiy]

There is no HTTP Data Provider!
Replase with ssh, etc

Also maybe instead of two identical Data Providers have one with ssh another s3 ?
Perhaps same with bourreau - the first one is slurm, the second Cloud Batch or 'any other HPC scheduler: ...'

[MontrealSergiy]

btw ftp is not secure, nothing to brag about even if it supported. Maybe best mention datalad, squashfs, or at least sftp

[prioux]

@MontrealSergiy We don't use FTP anywhere. We use SFTP.

[MontrealSergiy]

Therefore ftp should not appear on the diagram