When would you use AWS Glue versus Amazon EMR?

AWS Glue and Amazon EMR are both managed data‑processing services, but they target different use cases. Glue is a serverless ETL platform that abstracts cluster management; it automatically provisions resources, scales on demand, and charges only for the compute time your jobs consume. Glue is ideal for lightweight, scheduled ETL pipelines, schema discovery with crawlers, and integration with Glue Data Catalog. EMR, on the other hand, gives you a full Hadoop/Spark cluster that you can configure with custom AMIs, libraries, and cluster‑level settings. EMR is better suited for large‑scale, long‑running jobs, complex transformations, or when you need fine‑grained control over the cluster (e.g., specific Spark versions, YARN settings, or custom Hadoop components). In practice, choose Glue when you want quick, low‑maintenance ETL and can tolerate the Glue job runtime limits; choose EMR when you need performance tuning, custom dependencies, or integration with other big‑data tools. Both services can write to the same data lake, but Glue’s serverless nature often reduces operational overhead.