Refocus project as a reproducible PhD research artifact

[drona23]

Purpose

Refocus the repository from a product demo into a research-only artifact for carbon and water aware scheduling of deadline-flexible compute jobs.

Research question

Can a scheduler reduce operational carbon emissions and onsite water use by shifting non-preemptive jobs across time and location while preserving deadlines and hourly data center capacity?

What changed

• Removed the React frontend, API layer, Docker deployment files, and stale product-oriented code.
• Rebuilt the scheduler around explicit UTC handling, full execution-window accounting, deadline checks, and hourly capacity constraints.
• Added immediate-local, carbon-greedy, and seeded-random feasible baselines.
• Added paired-job comparisons so a policy cannot look cleaner by completing fewer jobs.
• Kept physical water separate from scarcity-weighted water impact.
• Added deterministic synthetic experiments, forecast-error stress tests, seed-level CSVs, metadata, figures, and SHA-256 checksums.
• Corrected the EIA fuel parser and ENTSO-E production-type and sub-hourly interval handling.
• Added CI, pinned dependencies, contribution guidance, citation metadata, BibTeX, and a 16-test suite.
• Rewrote the research problem, business beneficiaries, methodology, related work, data provenance, results, threats to validity, and reproducibility documentation.
• Removed all em dashes and en dashes from the repository text.

Result snapshot

Across 10 synthetic seeds with 180 jobs each, every policy completed all submitted jobs.

• Relative to immediate-local, the proposed joint policy used 23.94% less operational CO2e and 28.60% less physical water.
• Relative to random feasible placement, it used 22.31% less operational CO2e and 29.50% less physical water.
• Relative to carbon-greedy placement, it used 46.88% more operational CO2e and 52.75% less physical water.

The negative carbon result against the carbon-greedy policy is reported intentionally. It demonstrates the joint objective's carbon-water tradeoff and prevents a claim of universal superiority.

Validation

• Clean Python 3.11 environment from pinned requirements
• pip check: no broken requirements
• ruff check src tests: passed
• pytest: 16 passed
• Full benchmark regenerated
• SHA-256 manifest verified
• Relative Markdown links resolved
• CITATION.cff parsed
• git diff --check: passed
• Repository-wide em dash and en dash scan: passed
• All three committed figures visually inspected

Scope

The reported results are simulation evidence, not measured savings for a real organization. The artifact does not model embodied emissions, electricity-generation water, marginal emissions, data sovereignty, hardware compatibility, job preemption, or production failure modes.

Owner decisions before application or publication

• Replace the placeholder contributor in CITATION.cff with the preferred author name and ORCID.
• Select a software license. No license was added because that is a legal choice for the repository owner.