Source-audited research on graph database architecture, correctness, latency, resource efficiency, distributed execution, object-storage economics, and reproducible benchmark qualification.
This repository asks a deliberately difficult systems question: what would an engine need to prove to serve very-low-latency graph queries over PB-scale and trillion-edge datasets while using bounded compute and S3-class durable storage? It audits existing engines for reusable ideas and evidence gaps; it does not turn vendor claims into measured facts.
- Pin released source, documentation, images, and benchmark artifacts.
- Separate shipped behavior from development branches.
- Label source facts, official statements, vendor claims, issue reports, local observations, inferences, and unknowns.
- Treat exact result semantics and durability as benchmark prerequisites.
- Account for clients, caches, replicas, indexes, background work, recovery, and operator time when comparing resources or cost.
- Publish raw, reproducible artifacts before making a 10x claim.
docs/contains architecture specifications, the engine landscape, source audits, and benchmark protocols.src/contains corpus maintenance scripts and repository validation tools. Dedicated Aerospike research is maintained directly as reviewed Markdown rather than generated output.- CONTRIBUTING.md defines evidence and change standards.
Research is current to the date recorded in each specification. Database releases, pricing, source heads, issues, and managed-service behavior change; revalidate decision-critical claims before adopting them.
Start with:
- Research corpus index
- Cross-engine landscape
- Target architecture
- Reproducible benchmark and 10x protocol
Deep source audits:
The repository's original text and scripts are available under the MIT License. Linked third-party sources, product names, and quoted material remain subject to their respective owners and licenses.