TimescaleDB: Re-engineering PostgreSQL as a time-series database
summary
An open-source Postgres extension that exposes full SQL while reworking storage and inserts into hyper tables/chunks with automatic chunk creation and time + space partitioning. Claims ingest of millions of rows per second and handling hundreds of billions of rows, up to ~20x faster than vanilla Postgres and orders-of-magnitude faster than some time-series DBs for complex group-by queries. Implements fast retention by dropping chunks, runtime chunk exclusion, planned scale-out and continuous aggregates for automated downsampling.
AI-written summary · about
chapters (21)
- 0:00Start
- 3:42Existing databases don't work for time series
- 4:37Simplify your stack
- 6:12B-tree Insert Pain
- 7:01Challenge in scaling up
- 7:15TimescaleDB Scalable time series database, full SOL
- 15:32Multi-dimensional partitioning
- 16:30Automatic Space-time Partitioning
- 17:01Auto-tuning partition sizes
- 21:27Performance of TimescaleDB vs. PostgreSQL 10
- 23:25TimescaleDB: Repartitioning is Simple
- 24:36Conflict resolution
- 26:05Ellicient retention policies
- 26:51Single node: Scaing up via adding disks
- 28:23Multi-node: High availability and scaling read throughput
- 29:17Avoid querying chunks via constraint exclusion
- 31:12Additional time-based query optimizations
- 32:59Optimized for many chunks
- 34:05Rich Time Analytics
- 36:36Continuous aggregations and hierarchical views
- 37:13Multi-node: Scaling out across shared primaries
related talks
51mBuilding a scalable time-series database on PostgreSQL
51mPostgreSQL, The Time-Series Database You Want
42mAccessible Time Series data
13mTransform your business with time-series data (PRT061)
49mPostgreSQL, The Time-Series Database You Actually Want
53mEfficient Time Series with PostgreSQL
29mCatching Waves with Time-Series Data
58mIntroduction to Timeseries Databases