Does Druid work for non-timeseries data?

You probably already know that:

Time-optimized partitioning

Druid intelligently partitions data based on time and time-based queries are significantly faster than traditional databases.

But some people have used Druid for non-timeseries data. Here’s an interesting use case.

The TL;DR version:

The dataset has 200 million records, where each row represents a set of facts. Since there aren’t any observations over time, each row can be thought of as a “state.” Druid was selected due to its aggregation speeds, and the user was “blown away” by the aggregation speeds.