Explain like I'm five · Database

What is NoSQL Types & Modeling?

The four NoSQL types, told as a story about a warehouse that reshapes its shelves to fit each box.

the shelf slots are all the same shapebut the boxes arriving all have different shapeslong boxes, round boxes, bundled boxes...

The boxes arriving all have different shapes.Long boxes, round boxes, bundled boxes...

forcing boxes into fixed slots leaves one empty and one overflowingthe fixed slot shape keeps new box shapes from fitting

Forcing boxes into fixed slots leaves one empty and one overflowing.The fixed slot shape keeps new box shapes from fitting.

xyNoSQL = a warehouse that reshapes its shelves to fit the boxnot one fixed slot — many different ways to store boxes

NoSQL is a warehouse that reshapes its shelves to fit the box.Not one fixed slot — many different ways to store boxes.

Key-Valuefetch straight by tag
Documentfree-form inside the box
Column-familystores grouped columns under a row key
Graphboxes linked by string
four NoSQL typeskey-value — fetch by tag directlye.g. cachedocument — free-form inside the boxe.g. product infocolumn-family — columns grouped under a row keye.g. logsgraph — boxes linked by stringe.g. friend networksNoSQL stores boxes differently from the relational (normalized) warehouse

NoSQL splits into four types.NoSQL stores boxes differently from the relational (normalized) warehouse.

boxes can go in without a preset shapemany products are weak on multi-box receipts (transactions)if every ACID promise must hold, a relational warehouse is the usual pick

NoSQL is flexible, but many products have weaker support for transactions binding many boxes.If every ACID promise must hold, a relational warehouse is the usual pick.

In one breath

NoSQL is a warehouse that stores data in whatever shape fits — key-value, document, column-family, or graph — instead of fixed table slots.

NoSQL ("Not Only SQL") is an umbrella term for data models that handle large, unstructured data flexibly without a fixed schema. It favors horizontal scaling (sharding) but typically supports ACID transactions more narrowly than relational DBs. Modeling starts from query patterns and denormalizes data to fit them.

When grown-ups say it

Key-Value
One tag, one box. The simplest structure, and fast when looking up by key.
Document
Free-form content inside the box. Holds nested structure like JSON.
Column-family
A wide shelf holding column groups under each row key. Rows can have different columns. Unlike an analytical columnar DB, it doesn't store each column separately.
Graph
Boxes linked by string. Strong at tracing relationships.
Normalization
The rule of storing without duplication on fixed shelves. NoSQL applies this rule loosely. → the rule of writing it once
Sharding/Partitioning
Splitting the warehouse into zones. NoSQL usually supports sharding easily. → the warehouse split into zones
ACID
The four promises a receipt keeps. NoSQL may not keep ACID as strictly as the relational warehouse. → the four promises a receipt keeps
In-Memory DB
A desk that skips the deep shelf and serves straight from memory. Some products, like Redis, are both key-value NoSQL and in-memory. → the express pickup desk
Next story
Splitting an Overgrown Warehouse into Zones →
Want another term as a picture book? Request →