100x faster data app experiences delivered cost-effectively
summary
Firebolt claims to deliver sub-second analytics for data applications using column indexes, a proprietary TripleF file format, and server engines tuned by CPU/RAM sizing. A live demo runs on a $4 engine against a 100+ billion row, 11 TB fact table and repeatedly scans 15 MB from 11 TB to answer joins and text filters in milliseconds. Pricing is flat by engine time not queries and SQL lambdas handle semi structured arrays, making the system both fast and cost predictable.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:47Data applications
- 1:28Compromise
- 1:48Firebolt
- 2:35Similar Web
- 3:20Ranking Distribution
- 5:32What were we all about
- 6:55Choosing the right compute
- 8:14Choosing the engine scale
- 8:41Using indexes
- 9:09Semistructured data
- 9:46Demo
related talks
01h 04mHow We Build Firebolt
01h 08mFirebolt: Why Powering User Facing Applications on Iceberg is Hard
30mParticle physics, 10,000 times faster
19mDo You Believe in Magic?
26mColumnar Storage: Redefining Data Management for the Modern Era
17mHow Similarweb Serves 100s of TBs to Their Worldwide Users in Milliseconds
47mGábor Szárnyas | Data Track | Compass AI & Data Summit 2025
40mDuckDB: Harnessing in-process analytics for data science and beyond