# NoSQL Database

A broad family of non-relational database systems designed around models such as key-value, document, wide-column, or graph data.

- Type: term
- Domain: Engineering & architecture
- Level: Intermediate
- Last reviewed: 2026-09-23
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/nosql-database/>)

## What it means

NoSQL is an umbrella term for databases whose primary model is not the traditional relational table model. Major families include key-value, document, wide-column, and graph databases, each with different query, consistency, transaction, and scaling behavior. NoSQL does not mean schemaless or that SQL is never supported.

## Why it matters

A specialized model can make a dominant access pattern simpler, faster, or easier to distribute than a general relational design. The choice also moves responsibilities for relationships, constraints, consistency, and querying, so teams must evaluate a particular product rather than the label.

## When to use it

Consider a NoSQL database when a clearly understood workload benefits from aggregate documents, direct key access, graph traversal, sparse wide rows, or product-specific distribution characteristics. Prefer a relational database when joins, integrity constraints, and multi-record transactions dominate and no measured need justifies another model.

## How it works

1. Define the workload, access patterns, data relationships, consistency needs, scale, and failure expectations.
2. Choose the NoSQL family and product whose native model and guarantees match those requirements.
3. Design keys, partitions, documents, edges, or column families around the required reads and writes.
4. Specify validation, concurrency, transaction, retention, and secondary-index behavior explicitly.
5. Test representative data volume, skew, failures, resharding, backups, and recovery.
6. Monitor hot partitions, latency, capacity, consistency anomalies, and model evolution in production.

## A practical example

A content product stores each published article and its embedded blocks as one document because they are read and versioned together. It keeps billing records in a relational database, avoiding a single-store rule and choosing each model from its consistency and access needs.

## Where it breaks

The category contains systems with incompatible models and guarantees, so general claims about NoSQL are unreliable. Denormalization can duplicate data, cross-record operations may be harder, query flexibility can narrow, and distribution introduces operational and consistency costs.

## Common mistakes

- Choosing NoSQL because the data is JSON without analyzing relationships and transactions.
- Calling a database schemaless while leaving implicit structure unvalidated and undocumented.
- Applying relational modeling habits unchanged to a model designed around access patterns.
- Assuming every NoSQL product provides the same consistency, query, and scaling behavior.

## Visualization

[View the visualization on the website](<https://wiki.soroush.uk/topic/nosql-database/#visualization-nosql-database>)

## Sources

1. [What Is a NoSQL Database?](<https://aws.amazon.com/nosql/>) — Amazon Web Services · Official technical overview
2. [Databases and Collections](<https://www.mongodb.com/docs/manual/core/databases-and-collections/>) — MongoDB · Official technical documentation

## Related topics

- [Database](<https://wiki.soroush.uk/topic/database/>)
- [Relational Database](<https://wiki.soroush.uk/topic/relational-database/>)
- [Database Schema](<https://wiki.soroush.uk/topic/database-schema/>)
