S3: Object storage. The foundation. Store raw files, Parquet, Delta tables. Cheap, durable, scalable.
Glue: Serverless ETL. Crawlers detect schemas. Spark jobs run without cluster management. Data Catalog stores metadata.
Athena: Query S3 with SQL. Pay per query. Great for ad-hoc analysis. Uses Trino (formerly PrestoSQL) as its query engine.
EMR: Managed Hadoop/Spark. For heavy processing workloads. More control than Glue but more management overhead.
Redshift: Data warehouse. Columnar storage. Good for BI workloads. Redshift Serverless reduces ops burden.