Wide-column databases are built for very large datasets spread across many machines, especially when you already know how the application will read the data.

They are common in systems that store event logs, activity feeds, telemetry, audit records, time-series data, messages, and other workloads that write a lot. The usual pattern is simple: write a huge amount of data, spread it across many machines, and read it back using a carefully chosen key.

The name can be misleading. A wide-column database is not just a relational table with hundreds of columns. It is a storage model where data is organized around row keys, partitions, and groups of related columns.

Different rows may contain different columns, and the data is laid out on disk for fast key-based access at scale.

The trade-off is important: wide-column databases scale very well, but they expect you to design tables around the queries you will run. They are a poor fit when the application needs lots of joins, flexible filtering, or complex transactions.

This chapter covers the wide-column model and where it scales well.