Realtime Analytics with StarRocks in Wellness Booking
summary
Fresha replaces a brittle Postgres + Snowflake mix with StarRocks sitting between Iceberg lake tables and Kafka/Flink CDC, handling half a million bookings per day and billions of CDC rows. The system delivers subsecond OLAP: P99.9 queries drop from ~12–15s to ~300ms and P95 sits under 80ms, with one‑second staleness and one‑minute freshness tiers. Shared‑data mode, MySQL protocol, federation to Elasticsearch and caching cut compute and dev cost (one off autoscale bug spikes cost then remedied to ~$15–20/day).
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:24Traditional reporting limits
- 7:42Partner needs for real-time
- 13:03Introducing StarRocks
- 23:43StarRocks key features
- 27:10Homepage analytics case
- 30:23Payment lock case
- 32:10Data platform roadmap
- 34:15Q&A
related talks
18mOpen-Source OLAP on Open-Source Data Warehouses.
20mEnabling Machine Learning through Real-Time Data Processing
32mDiscover the Modern Way to Build Real-Time Analytics Solutions at Any Scale
47mBuilding a Real-Time Analytics Database
08mReal-Time Data Systems: Speed, Scale, and Resilience
34mReal time OLAP and Stream Processing Friends or Foes
29mAccelerating Performance: Mastering RocksDB for High Speed Data Processing
32mBuilding Real-time Systems with Open Source Technologies