Blazing-Fast Python in Your Database: Unlocking Data Science at Scale with Exasol
summary
Exasol runs native Python UDFs inside an MPP database using script language containers, a bucket file system for distributing models, and SQL orchestration to parallelize work; a telecom case builds ~120,000 separate time-series models (one per tower) with the Prophet library and executes them in parallel, cutting a multi-week job down to days and keeping models and predictions inside the database for immediate SQL-based analytics. 120,000 towers Prophet library reduced weeks to days
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:13Data science scaling challenges
- 2:37Exasol high-performance database
- 4:02Python in SQL with UDF
- 7:54AI and ML architecture
- 8:58Telecom provider use case
- 11:25Deploying Python models in SQL
- 13:53In-database AI benefits
- 16:40Customer success story
- 17:21Q&A: Knowledge graphs
- 18:33Q&A: Geospatial data
- 20:39Q&A: Performance benchmarks
- 22:22Q&A: Inference time metrics
- 23:46Q&A: TPC benchmark dominance
- 25:52Q&A: Open source licensing
- 26:34Q&A: MPP and UDF distribution
- 27:38Closing remarks
related talks
37mAccelerating Python Analytics by In-Database Processing
26mPython + MPP Database = Large Scale AI/ML Projects in Production Faster
41mMassively Parallel Processing with Procedural Python
01h 06mBodo: Automatic HPC Performance and Scaling for Data Processing in Python
25mProcessing Biggish Data with DuckDB and Python
30mScaling Outside the Warehouse Using DuckDB and Python
39mAccelerating Data Science with RAPIDS
39mPython and SQL: Better Together, Powered by DuckDB