Your data lakes need architecture too!
summary
Covers concrete architectures and tradeoffs: the original two-layer architecture (landing + raw) for cheap cold storage, Bill Inmon’s Data Pond split (application / textual / analog) that forbids reprocessing, multi-layer and zonal designs that enforce Medallion (bronze/silver/gold) flows and security, and lakehouse variants (Delta Lake, Databricks, Snowflake) that add ACID, time travel and require ETL/ELT pipelines. Notes common failures: unmanaged lakes turn into swamps, lakehouses incur added tooling, licensing and operational cost; recommends Airflow, Spark, Hadoop/Elastic and clear governance for reusability.
AI-written summary · about
chapters (10)
- 0:00Data lake architectures
- 1:12Two-layered architecture
- 4:22Inmon's data ponds
- 8:02Multi-layered architecture
- 11:41Zaloni's zonal architecture
- 14:40Zonal architecture with Data Vault
- 16:04Data lakehouse explained
- 20:40Medallion architecture layers
- 22:06Usage scenarios and conclusion
- 23:10Q&A
related talks
31mData Warehouses Meet Data Lakes
44mWhat's the Best Big Data Architecture for You?
50mBehind the Hype - Architecture Trends in Data
51mHow to Build a Modern Data Lakehouse?
21mWhy the Lakehouse?
41mOn Data: Sky above, sand below, peace within
48mAn Evolution of Data Architectures - Lambda, Kappa, Delta, Mesh & Fabric
41mModern Data Architectures – What They Are and How to Implement?