AWS Glue
Serverless ETL and the Data Catalog for your lake.
❓ What is it?
A serverless data-integration service: crawlers discover schemas into a central Data Catalog, Spark/Python jobs transform data at scale, and DataBrew adds a visual no-code preparation layer.
💡 Why does it exist?
Models eat clean, well-shaped data; raw sources are anything but. Glue turns scattered files and tables into catalogued, query-ready datasets without a cluster to manage.
⏱️ When should you use it?
Use Glue when data must be cleaned, joined, converted (e.g. CSV→Parquet), or catalogued before training or analytics; use DataBrew when analysts need visual transformations instead of code.
🗺️ Where does it fit?
Between raw sources and consumers: it crawls S3/JDBC sources, writes transformed data back to S3, and its Catalog serves as the shared schema registry that Athena and SageMaker rely on.
🔌 How do you integrate it?
Run a crawler to populate the Catalog, author a job (visual or script), schedule with triggers/workflows, and monitor runs; call the Catalog from Athena queries automatically.
🧩 Commonly integrated with
🎯 Exam angle (AIF-C01)
- Glue = serverless ETL + Data Catalog. "Prepare/transform data for ML without managing servers" → Glue.
- DataBrew vs Data Wrangler: both visual prep — DataBrew is Glue/analytics-oriented, Data Wrangler lives inside SageMaker for ML.