The Data Lakehouse
summary
Argues that a data lake becomes usable only after adding a searchable analytical infrastructure = metadata layer (taxonomies, ontologies, inline contextualization), plus cleansing on ingest: textual ETL for text and distillation of analog IoT that separates the few high-value events into the lake and pushes bulk low-value streams to cheap storage. Claims this converts lakes into actionable lakehouses, benefits analysts, and is a long ongoing engineering project rather than a one-off.
AI-written summary · about
chapters (18)
- 0:00Future of data architecture
- 1:28Corporate data types
- 3:19Textual data processing challenges
- 5:27Data lakes to lakehouses
- 6:33Building analytical infrastructure
- 8:44Analytical infrastructure as catalog
- 10:13Metadata beyond models
- 11:37Data cleansing for text
- 13:04Analog data distillation
- 15:36Managing data volumes
- 17:25Bulk storage and archiving
- 18:08Benefits for data teams
- 19:16Lakehouse as ongoing project
- 19:40Warehouse versus lakehouse
- 21:30Free lakehouse resources
- 23:36Q&A on lakehouse and AI
- 25:48AI needs clean data
- 29:06Final Q&A and closing
related talks
34mData Lakehouse: Who, Why and How
31mData Warehouses Meet Data Lakes
26mUnifying Data for Data Applications with the Data Lakehouse
29mYou Have A Data Lake, Now What?
02h 05mBig Data and the Data Lake from the Analyst Perspective
04mInterview
25mBuild open format data lakehouse
11mBreaking Data Silos: Enterprise Data Management