Catching Waves with Time-Series Data
summary
Shows concrete numbers and designs: benchmarked claims like TimeScaleDB loads 1B rows in 1/15th the time of Postgres, query latency 396× faster on 100M rows, and compression "up to 90%". Describes a built internal system LittleTable stores 320 TB (2017) and uses hierarchically delineated keys. Explains exact storage patterns (column vs row), concrete aggregation strategy (base table + day/week/month aggregates with TTLs like 1 month / 6 months / 1 year / 5 years), and an API contract that requires prefixing the composite key (must include region if providing break).
AI-written summary · about
chapters (19)
- 0:00Time-series data intro
- 1:05What makes time-series unique
- 2:30Choosing time-series tools
- 4:58Postgres extensions vs vanilla
- 6:43Dedicated TSDBs compared
- 8:06Building your own TSDB
- 9:32Efficient TS data organization
- 10:17Indexing and composite keys
- 12:23Hierarchical key design
- 14:28Querying TS data effectively
- 15:55ClickHouse composite key quirks
- 17:17Data retention and compression
- 19:24Columnar storage and compression
- 21:36Aggregation for long-term storage
- 23:40TS API endpoints design
- 26:45Visualizing TS trends
- 27:05Postgres limits and integrations
- 28:07Real-world TS insights
- 28:49Closing and Cisco Meraki
related talks
30mCatching Waves with Time-Series Data
51mPostgreSQL, The Time-Series Database You Want
49mPostgreSQL, The Time-Series Database You Actually Want
39mTimescaleDB: Re-engineering PostgreSQL as a time-series database
51mBuilding a scalable time-series database on PostgreSQL
58mIntroduction to Timeseries Databases
13mTransform your business with time-series data (PRT061)
42mAccessible Time Series data