Glue, DataBrew, EMR, Data Wrangler: picking the tool
Lesson 1 of 5 in Transforming Data and Engineering Features.
Four AWS tools all claim to "transform data," and telling them apart is one of the most reliably tested forks in Domain 1. The exam separates them by persona (who is doing the work), interface (code or visual), and infrastructure control (serverless or a cluster you shape). AWS Glue is serverless Spark ETL for data engineers. Glue DataBrew is point-and-click preparation for analysts who write no code. Amazon EMR hands you managed Spark/Hadoop clusters when you need control, custom libraries, or an existing big-data estate. SageMaker Data Wrangler is the ML-native option inside SageMaker Studio, built to end in a training set, a Pipeline step, or a Feature Store ingestion.
| Tool | Interface & persona | Infrastructure | Reach for it when |
|---|---|---|---|
AWS Glue | PySpark/Scala code, or visual DAGs in Glue Studio — data engineers | Serverless Spark; pay per job | Production ETL on a schedule, Catalog-integrated, least ops for Spark |
Glue DataBrew | Point-and-click recipes, 250+ built-in transforms — analysts, zero code | Fully managed | Visual cleanup and profiling by people who will not write Spark |
Amazon EMR | Your Spark/Hive/Presto code — big-data engineers | Clusters you size and tune (or EMR Serverless) | Custom frameworks and libraries, very large jobs, lift-and-shift from on-prem Hadoop |
SageMaker Data Wrangler | Visual, ML-centric, inside Studio — data scientists | Managed by SageMaker | Feature prep with built-in data-quality and bias insights; exports to Pipelines, Feature Store, or a notebook |
Pick the transform tool
Interactive decision tree — outcomes:
- AWS Glue
Serverless Spark with Data Catalog integration — the default for coded, scheduled, production ETL with the least infrastructure to run.
- Amazon EMR
Managed clusters with full control: your library versions, your tuning, any ecosystem engine. The price is that you shape and manage the cluster.
- SageMaker Data Wrangler
Visual preparation that speaks ML: quality and bias reports built in, and one-click export to SageMaker Pipelines, Feature Store, or notebook code.
- AWS Glue DataBrew
No-code recipes and profiling for analysts. Repeatable, auditable cleanup without a line of Spark.
Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.