← Lessons

quiz vs the machine

Silver1050

System Design

The Time Series Data Model

How metrics are shaped as named streams of timestamped numbers.

4 min read · intro · beat Silver to climb

What a time series is

A time series is an ordered sequence of points, each a pair of a timestamp and a numeric value. Time series databases are built around the fact that data arrives in time order and is almost always queried by time range.

Identifying a series

A single series is identified by a metric name plus a set of labels (key value pairs). Together they form a unique identity. For example a metric named http requests total with labels for method and status code is one series, while a different status code is a different series.

Why this shape matters

  • Points for one series are append only and roughly sorted by time, which enables tight compression.
  • Queries select series by name and labels, then scan a time window.
  • The value is usually a float, sometimes an integer counter.

The cardinality consequence

Every unique combination of labels creates a new series. This makes the model expressive but also dangerous, since careless labels multiply series count.

Key idea

A metric is not one number but a family of series, each pinned by a name and labels, storing ordered timestamp value points.

Check yourself

Answer to earn rating on the learn ladder.

1. What uniquely identifies a single time series?

2. Why does the time ordered append only shape help storage?