NoSQL

New
assess
First Added:May 28, 2026 Updated: June 12, 2026

NoSQL

NoSQL (“Not Only SQL”) names data stores that model and scale outside classic relational tables. Document, key-value, wide-column, and graph systems each trade different consistency and query patterns. We assess the family when requirements clearly need that shape; default to Postgres (adopt) until relational limits show up in production.

Blurb

NoSQL databases are designed to handle large volumes of unstructured, semi-structured, or structured data, and they often prioritize scalability, flexibility, and performance over strict consistency.

Summary

What it is: A design space, not one product. Examples include MongoDB (document), Redis (key-value), Cassandra (wide-column), and Neo4j (graph). usql and BI tools may connect to several of these alongside SQL engines.

When to use: Schema churn or nested documents dominate; extreme horizontal write scaling; TTL-heavy caches; graph traversals are the core query pattern.

When to skip: Most business apps with joins, transactions, and reporting on a stable schema (Postgres). Analytics warehouses (BigQuery) when SQL over huge columnar data is the real need, not document flexibility.

Key tradeoffs: Eventual consistency models, operator skill split, and operational sprawl when every service picks a different engine.

Details

StyleTypical useRelational alternative
DocumentCMS, catalogsPostgres JSONB
Key-valueSession cacheRedis with Postgres source of truth
Wide-columnHigh-ingest time seriesPartitioned Postgres or warehouse
GraphIdentity graphsPostgres recursive CTEs or dedicated graph DB

Practices: Prove you outgrew Postgres JSONB or replicas before adding a second datastore; standardize on one document store per org if you must; keep Redash queries read-only.

References