Some Like It Hot
summary
Explains why Apache Druid becomes a hot serving layer for interactive analytics by ingesting and indexing data ahead of queries, using columnar segments (a few million rows each), dictionary encoding and roaring bitmaps to operate on compressed data, and fanning queries across thousands of segments. Claims real deployments ingest 100+ billion rows per day, deliver sub-second query latency, and show 20–59× speedup over Presto (100× over Hive) while trading off ingestion/partitioning work for runtime performance.
AI-written summary · about
chapters (13)
- 0:00Intro
- 6:57Data Warehouses
- 14:21Druid
- 15:15Use Cases
- 16:06Benchmarks
- 17:03Architecture
- 18:10Parallelization
- 19:15Druid Segment
- 21:55Druid Query
- 22:37Filter
- 24:17Aeration
- 25:42Read
- 26:19Recap
related talks
28mThe Heat is On: Architecting for Hot Analytics
39mApache Druid 101
27mTales at Scale: Analytics at 1000 QPS and Beyond
27mSucceeding with Apache Druid and Clickstream Data
28mBest of Stream + Batch: When Streaming Analytics Isn’t
21mPutting big data self-service analytics in the hands of business people
20mAnalyzing streaming data with Apache Druid
26mColumnar Storage: Redefining Data Management for the Modern Era