Skip to content

Research_Computing_Platform_Architecture_Handbook.md

Latest commit

 

History

History
108 lines (70 loc) · 2.53 KB

Research_Computing_Platform_Architecture_Handbook.md

File metadata and controls

108 lines (70 loc) · 2.53 KB
Raw

Research Computing Platform Architecture & Engineering Handbook

Table of Contents

  1. Vision
  2. Guiding Principles
  3. Target Architecture
  4. Repository Standards
  5. Data Standards
  6. Shared Libraries
  7. Configuration
  8. Testing & Release Discipline
  9. AI Development Workflow
  10. Migration Roadmap
  11. Timeline
  12. Governance

1. Vision

Create a project-agnostic engineering platform where code, data, configuration, and runtime are independent. Individual repositories continue evolving without disruptive platform rewrites.

2. Guiding Principles

  • Code in Git; data outside Git.
  • One source of truth for shared data.
  • Shared utilities live in one core package.
  • Projects remain independently deployable.
  • Every platform change is reversible.

3. Target Architecture

Development/
Data/
Shared/
Workspaces/
Documentation/

4. Repository Standards

Standard layout, .env, tests, docs, CI, no hard-coded paths.

5. Data Standards

Single DATA_ROOT, logical resource names, versioned datasets, backups, manifests.

6. Shared Libraries

Introduce platform-core incrementally. Keep compatibility wrappers until migrations finish.

7. Configuration

Environment variables only; machine-specific values outside repositories.

8. Testing & Release Discipline

Baseline before every phase. Smoke tests. Regression tests. Rollback plan. No platform change without passing all tests.

9. AI Development Workflow

Use Copilot/ChatGPT against written implementation plans. Implement one phase at a time. Review diffs before merge.

10. Migration Roadmap

Phase 0 (1 week)

Inventory repositories, dependencies, duplicated utilities. Exit: no code changes.

Phase 1 (1-2 weeks)

Centralize configuration and paths. Exit: applications unchanged.

Phase 2 (2-4 weeks)

Shared data layer and manifests. Migrate one pilot repository.

Phase 3 (2-6 weeks)

Extract shared platform-core package. Migrate repositories gradually while active feature work continues.

Phase 4 (ongoing)

Standardize new repositories on platform template.

Phase 5 (future)

Cloud readiness and automation.

11. Timeline

Months 1-2: foundations. Months 3-4: migrate highest-value repositories. Months 5+: continuous modernization while feature development proceeds.

12. Governance

Treat this handbook as the engineering standard. Update by version. Every architectural change requires documented rationale, migration path, and rollback plan.