SQL vs NoSQL

SQL vs NoSQL

Step 1 of 12

The same invoice, two shapes. The question is which side pays: writes, or reads.

The Idea

A relational database enforces a schema, stores each fact once, and joins tables at query time with transactional guarantees. A document store keeps whole objects together, lets fields vary between records, and expects you to duplicate data so a read needs no join. The question is which side pays: writes, or reads.

Real-World Example

A garage runs both. Invoices live in strict columns — parts, labour, tax — because the totals must reconcile and nothing may be half-written. Inspection notes do not: a motorcycle check has fields a van check never heard of, and nobody wants a schema change per vehicle type.

The Tradeoff

Schemaless does not mean structure-free; the structure just moves into application code, where nothing enforces it. Duplicated data means every copy needs updating on write. Reach for documents when a shape genuinely varies and is read whole — not as a way to avoid designing the data.

Your turn

Put the steps in the right order.

1. 1Check whether that shape is stable across every record↑↓
2. 2Pick the store whose cost lands where you still have headroom↑↓
3. 3List the queries this feature actually has to answer↑↓
4. 4Decide whether a write must be all-or-nothing across entities↑↓

Check order

Mini quiz

1 / 3

Storing each fact once and joining at read time describes:

• AA relational database
• BA document store
• CA blob store
• DA message queue