Continuous Aggregates A Use Case of Materialized Views
summary
TimescaleDB implements continuous aggregates that incrementally refresh materialized results by tracking a watermark and applying aggregation only to newly inserted rows, cutting work compared with full REFRESH MATERIALIZED VIEW. A background scheduler runs refresh policies; out‑of‑order inserts produce entries in an invalidation log so only affected time ranges re‑aggregate. A real‑time view unions materialized buckets before the watermark with raw rows after it. Hierarchical continuous aggregates (since 2.9) let minute→hour→day layers speed initial refreshes; first full populate is single‑threaded and can be slow.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:01What is TimescaleDB?
- 3:20PostgreSQL: setup
- 5:57PostgreSQL: Materialized Views
- 9:29TimescaleDB: Continuous Aggregates
- 21:27Continuous Aggregates: Limitations
- 22:29Continuous Aggregates: Future
- 25:16Questions
- 30:30TimescaleDB: Hierarchical Continuous Aggregates
related talks
02h 00mTutorial
01h 05mIncremental View Maintenance 2
58mIntroduction to Timeseries Databases
45mpg_ivm - Extensions for Rapid Materialized View Update
38mIncremental View Maintenance: Beyond Worst-Case Analysis I
49mIncremental Computation
06h 54mTopical Edition on Databases - Track 1
38mTime-Based Aggregates: The Good, the Bad and the Ugly