Overview

Serverless query service to analyze data directly in S3 using standard SQL.
Built on Presto engine.
No infrastructure to manage, no data loading — data stays in S3.
QuickSight can connect to Athena for dashboards and data visualization.

Pricing

Pay per query, based on the amount of data scanned (~$5 per TB scanned).
Less data scanned = cheaper and faster → this drives all the performance tips.

Performance Improvement

• Use columnar formats — Parquet or ORC (instead of CSV/JSON) so only needed columns are scanned. Use Glue to convert to these formats.
• Compress data — bzip2, gzip, lz4, snappy, zstd, etc. to reduce data scanned.
• Partition datasets — organize data in S3 by columns (e.g. s3://bucket/year/month/day/) so queries only scan relevant partitions.
• Use larger files — bigger files (>128MB) minimize overhead vs many small files.

Federated Query

• Run SQL queries across data sources beyond S3 — both relational and non-relational, AWS and on-premises.
• Uses Data Source Connectors that run on AWS Lambda to execute the federated queries.
• Example sources: RDS, DynamoDB, Redshift, CloudWatch Logs, ElastiCache, on-prem databases.
• Query results can be stored back in S3.

Athena vs Redshift

• Athena is fully serverless; data stays in S3 (good for ad-hoc queries on S3 data).
• Redshift is a provisioned/serverless data warehouse, faster for complex repeated queries thanks to indexing and pre-loaded data.
• See [[Redshift]].

Exam Tips

• "Analyze / query data in S3 with SQL, serverless" → Athena.
• "Reduce Athena cost / improve performance" → columnar format (Parquet/ORC) + compression + partitioning.
• Common combo: S3 + Athena + QuickSight = serverless analytics & BI.
• Query logs (VPC Flow Logs, ELB logs, CloudTrail) stored in S3 → Athena.