AWS Glue
- What it is
- Serverless ETL (extract, transform, load) service. Discovers, prepares, moves, and integrates data for analytics and ML.
- No infrastructure to provision; pay per DPU-hour of job run time.
- The parts
- Crawlers — scan a data store, infer schema and partitions, and register them in the AWS Glue Data Catalog.
- ETL jobs — Python or Spark scripts, generated visually or written by hand.
- Glue Studio — visual job authoring. Glue DataBrew — no-code visual data preparation for analysts.
- Workflows and triggers — orchestrate multi-job pipelines.
- Not to be confused with
- Amazon EMR — you manage a Hadoop/Spark cluster with full control over frameworks and tuning. Glue is serverless and opinionated.
- AWS DMS (Database Migration Service) — replicates databases; it moves data, it does not transform it into an analytics schema.
- Exam signal
- “Serverless ETL”, “prepare data for analytics”, “discover schema” → AWS Glue.