Database Engines Research Wiki

A cross-linked reference covering the top ~150 database engines (by db-engines.com ranking) plus the adjacent data-platform technologies that shape real database decisions (lakehouse table formats, streaming platforms, streaming/real-time databases, CDC, catalogs, query engines) and the theory concepts behind them.

Each engine page answers, quickly: what is it, how does it behave under stress, and when is it the wrong tool? — with light citations, explicit confidence levels, and honest ⚠️ unverified flags where claims couldn't be confirmed.

Start here

• Decision guide — "Which database should I use?" Four questions, then a branching tree to candidates with the key trade-off and the anti-pattern.
• Index — full catalog, grouped by data model, one line each.

Layout

Path	What's in it
engines/	One page per engine. Ranked db-engines and adjacent tech (the latter flagged adjacent: true, rank: n/a, with a category:).
concepts/	Shared theory — CAP/PACELC, isolation levels, ACID vs BASE, LSM vs B-tree, lakehouse, and more. Explained once, linked everywhere.
index.md	Catalog of every page.
decision-guide.md	The "which DB?" tree.
ranking.md	The worklist (top-150 + a separate adjacent-tech table).
CLAUDE.md	The schema: conventions and workflows this wiki is built and maintained under.
log.md	Append-only operations log.
tools/	Maintenance scripts (link conversion + link checker).

Coverage: 150 ranked engines + 19 adjacent technologies + 34 concept pages.

Conventions

• Cross-links are standard relative markdown links (GitHub-native). The wiki was originally authored with Obsidian [[wikilinks]] and converted via tools/wikilinks_to_md.py.
• Confidence is recorded per page (high/medium/low); specific unverifiable claims carry an inline ⚠️ unverified — prefix.
• Sourcing prefers primary sources (official docs, design papers, Jepsen) over marketing; CAP/PACELC, isolation, and Jepsen claims are cited.

Maintenance

python3 tools/check_links.py        # verify all internal links resolve; flag stray wikilinks
python3 tools/wikilinks_to_md.py    # (idempotent) convert any [[wikilinks]] to markdown links

This is an LLM-maintained wiki: the research, writing, and cross-linking are done by Claude under the schema in CLAUDE.md; a human curates scope and asks questions.

How this wiki was bootstrapped

The wiki was drafted by Claude Code (Opus 4.8) over three sessions, based on Andrej Karpathy's LLM Wiki methodology, 99% Claude starting from broad direction on intent, concerns, and methodology:

1. convo0.txt — Ask Claude to brainstorm the rubric: the set of concerns / dimensions every database engine should be evaluated against (data taxonomy, CAP placement, schema model, query language, hardware profile, licensing, etc.).
2. convo1.txt — Ask Claude to write CLAUDE.md (the schema + workflows) based on the prompt/rubric from previous step, to bootstrap the full wiki using Karpathy's knowledge-base methodology — a folder of cross-linked markdown the LLM writes and a human reads.
3. convo2.txt — Ask Claude to draft the wiki using a long-running workflow (cogitated for 1h 13m 49s in one stretch), then began a manual review pass, addressing issues and improvements.

⚠️ Caveat: content is AI-generated from web research and may contain errors or go stale (rankings and licenses change). Pages note their last_researched date. Verify against primary sources before making decisions that matter.

License

Documentation/content: CC BY 4.0. Scripts under tools/: MIT (see LICENSE).