Data terms, decoded

Quick reference

• CDC — Change data capture: read the database log, not the table. Every change becomes an event, and the warehouse sees it in seconds.
• ELT — Extract, load, transform: land the raw data first, then transform it in SQL inside the warehouse. The raw copy stays replayable.
• OLAP — Online analytical processing: column stores that scan a few columns across billions of rows. Built for sums, not for one order by id.
• Lakehouse — Object storage plus an open table format (Iceberg, Delta or Hudi) plus a catalogue. Cheap like a lake, transactional like a warehouse.
• Lineage — The graph of which table and column fed which. It turns 'this report is wrong' into a list of downstream tables.
• Cardinality — How many distinct values a column holds. user_id is high, status is low, and the estimate decides join order.

Explained

CDC

Change data capture reads the database log, the WAL in Postgres or the binlog in MySQL, instead of polling the table. Every insert, update and delete becomes an event.

ELT

Extract, load, transform. Land the raw data first, then transform it with SQL in the warehouse, so the raw copy stays replayable when the logic changes.

OLAP

Online analytical processing: column stores built to scan a few columns across billions of rows. A sum over a month is fast; fetching one order by id is not.

Lakehouse

Object storage plus an open table format (Iceberg, Delta or Hudi) plus a catalogue. Cheap like a lake, transactional like a warehouse.

Lineage

The graph of which table and which column fed which. In an incident it names the upstream cause and everything downstream.

Cardinality

The number of distinct values in a column. user_id is high, status is low, and the planner's estimate of both decides join order and index choice.

In one line

Check cardinality

SELECT count(DISTINCT country) FROM users