Aggregates are summaries of raw data for a period of time. Some examples of aggregates include the average stock price per day, the maximum CPU utilization per 5 minutes, or the number of visitors on a website per week.
Calculating aggregates on time-series data can be computationally intensive. There are a few different reasons for this:
- Aggregating large amounts of data often requires a lot of calculation time.
- Ingesting new data requires new aggregation calculations which can affect ingest rate and aggregation speed.
TimescaleDB's continuous aggregates solve both of these problems. Continuous aggregates are automatically refreshed materialized views that speed up query workloads for large amounts of data. They solve some of the main pain points with materialized views and home-grown aggregate solutions in a couple of ways.
First, TimescaleDB processes the aggregation calculations when the aggregate is created and then stores the aggregation results to minimize re-calculation when new raw data is added.
Second, TimescaleDB provides ongoing updates to continuous aggregate data with an automatic continuous aggregate refresh policy. This schedules an automatic job that re-calculates new data for a specific interval of time. Thus, the policy only recomputes the newest changes in the raw data rather then recomputing everything.
For more information on the benefits of continuous aggregates, see the Overview page on Continuous aggregates.
Materialized views in PostgreSQL are table-like objects within your database. For more information on materialized views, see the PostgreSQL documentation.
Follow this tutorial to create a continuous aggregate and continuous aggregate refresh policy:
You only get the full benefits of continuous aggregates by creating both the aggregate itself and its policy. Follow both sections to add the full value of aggregates to your time-series database.
Found an issue on this page?Report an issue!