The Joe Reis Show Egor Gryaznov - The "Non-Modern Data Stack", and Getting Out of Our Data Bubble
15 snips
Oct 4, 2023 Egor Gryaznov, data industry expert, chats about the 'Non-Modern Data Stack' and the importance of getting out of our data bubble. They discuss building data warehouses at Uber, contrasting big tech companies in Silicon Valley, selling non-modern data solutions, and the significance of observability for the non-modern data stack.
AI Snips
Chapters
Transcript
Episode notes
Model Before You Log Everything
- Design a data model before logging everything to avoid scale pain later. Use dimensional modeling, separate facts and dimensions, and log keys rather than duplicating attributes.
Migrating Queries Revealed Core Models
- Igor's team migrated many Postgres queries into a central warehouse and studied common query patterns. They created core models like fact_trip, dim_driver, and dim_rider to simplify repeated complex joins.
The Modern Data Stack Is A Bubble
- Silicon Valley practices form a bubble that doesn't represent most enterprises' reality. Igor realized many companies outside the bubble run legacy stacks and need different, pragmatic solutions.