Advanced HBase Schema Design
summary
Covers concrete HBase choices that trade write throughput for read latency: use composite row keys and move high-cardinality attributes left into the row key, avoid hotspotting with salting/subsharding, and pre-materialize duplicated rows for fast reads. Recommends keeping column families small (under ten), using bloom filters, compression and tuned block size, and exploiting HBase versions for time-series like OpenTSDB. Shows which schemas concentrate load and which distribute it across regions.
AI-written summary · about
chapters (14)
- 0:00Introduction and agenda
- 0:47What is HBase logically
- 2:53Cell versions in HBase
- 4:16Column families and regions
- 8:11Column family use cases
- 12:23Schema design basics
- 13:47Sorting and performance
- 15:30Why schema design matters
- 16:36Denormalization in HBase
- 18:03Pre-materializing data
- 19:30Nested entities in rows
- 20:13OpenTSDB example
- 21:19Avoiding hot spotting
- 21:40Schema design summary
related talks
44mOpenTSDB - Time Series Schema on Schemaless NoSQL
51mThe Lost Art of Database Design
46mThe Lost Art of Database Design
01h 02mDesign Good Schemas - Get a Better Database
55mApache HBase: crazy dances on the elephant back
48mDesign of data-intensive applications
34mAvoiding Big Data Antipatterns
41mReal-World NoSQL Schema Design #bbuzz