Amazon Athena
Serverless SQL directly over data in S3.
❓ What is it?
An interactive query service that runs standard SQL against files in S3 (CSV, JSON, Parquet, ORC) using the Glue Data Catalog for schemas — no cluster, pay per data scanned.
💡 Why does it exist?
Loading data into a warehouse just to answer a question is friction. Athena queries the lake in place, making exploratory analysis and ad-hoc feature checks a 30-second task.
⏱️ When should you use it?
Use it for ad-hoc exploration of training data, building datasets from the offline Feature Store, log analysis, and lightweight reporting feeds into QuickSight.
🗺️ Where does it fit?
On top of S3 + Glue Catalog: analysts and pipelines submit SQL; results land back in S3; QuickSight and notebooks consume them.
🔌 How do you integrate it?
Define/crawl a table schema, write SQL in the console or via JDBC/SDK, and cut cost by converting data to partitioned Parquet.
🧩 Commonly integrated with
🎯 Exam angle (AIF-C01)
- "Query data in S3 with SQL, serverless" → Athena. Columnar formats (Parquet) + partitioning = the cost-optimisation answer.
- Athena is how you assemble training sets from the Feature Store offline store.