Menu
ESC

Nhập từ khóa để tìm kiếm

↑↓ Di chuyển
Enter Mở
ESC Đóng

Đang tải...

Bài 53 — Data Lake vs Lakehouse vs Warehouse

Data-Driven Organization Bài 53/60

Lake vs Lakehouse vs Warehouse

Quick compare

AspectWarehouseLakeLakehouse
DataStructuredAnyAny
SchemaOn-writeOn-readOn-write (Delta/Iceberg)
Cost$$$$$$
QueryFast SQLSlow (Hive)Fast SQL
MLLimitedYesYes
ToolsSnowflake, BQ, RedshiftS3 + HiveDatabricks, Iceberg

Warehouse strengths

  • Mature, BI-friendly.
  • ACID transactions.
  • Strong governance.

Lake strengths

  • Cheap raw storage.
  • Multi-format (JSON, parquet, image, video).
  • ML training data.

Lakehouse promise

  • Best of both — cheap + ACID + SQL + ML.
  • Delta Lake (Databricks), Apache Iceberg, Apache Hudi format.

Choosing

  • Pure BI + structured data → Warehouse.
  • ML + unstructured + flexibility → Lakehouse.
  • Cheap archive + occasional query → Lake.

VN reality

  • Most start warehouse-first (Snowflake/BigQuery).
  • Add lake for ML data (S3 + parquet).
  • Lakehouse adoption emerging 2024-2026.

Anti-pattern

  • Build lake without clear use case → data swamp.
  • Use warehouse for raw event store → expensive.
  • Migration warehouse → lakehouse without ROI analysis.