Amazon Athena

by Amazon Web Services

A serverless SQL query service that reads data directly from Amazon S3. There is no cluster to provision: queries run on demand and are billed on the data they scan.

Core model

  • Serverless SQL on S3 — queries run against files in S3 (Parquet, CSV and other formats) with no infrastructure to manage
  • Pay per data scanned — the on-demand rate is per terabyte scanned, so the main cost levers are partitioning, columnar formats and filtering
  • Capacity reservations — predictable workloads can buy Data Processing Unit (DPU) capacity instead of paying per scan
  • Apache Iceberg tables — Iceberg tables are created with a table property and support time-travel queries
  • Free operations — DDL statements and failed queries are not billed, per AWS’s pricing model as summarised in secondary sources
  • Separate S3 charges — S3 storage, requests and query results are billed by S3 in addition to Athena

Position

Athena is AWS’s query-in-place option for a data lake. It sits alongside Amazon Redshift, which is a
warehouse with its own storage and clusters, and it competes with other query engines such as Trino
and dremio for lake queries.

  • Amazon Redshift
  • BigQuery — the Google Cloud equivalent for serverless analytics
  • Dosi — compiles semantic models to Athena SQL in its 0.2.0 release

Sources