AWS Athena

Overview

• Interactive Data Query Service
  • Cannot be used to modify data
• Allows to query data on S3 
• Based on Presto
• Uses schema-on-read

Model

• Table
  • EXTERNAL (data stored in S3)
• Partition
  • Up to 20K partitions per table
  • Typically split by date
  • Can use Lambda to automatically create partitions
• S3 data location
• Metadata
  • Uses Apache Hive DDL
  • Stored in AWS Glue if availalble in Region
• SerDe (how to interpret a row)

Query Results

• Stored in S3
  • Can use KMS for encryption

Pricing

• Pay for data scanned in S3
  • $5 per TB of data scanned
  • Minimum 10 MB per query
• Optimizations
  • Compression (gzip)
  • Parititioning
  • Columnar data formats (e.g. Parquet)

References

• https://www.thomashenson.com/schema-read-vs-schema-write-explained/
• http://www.marklogic.com/blog/schema-on-read-vs-schema-on-write/
• https://www.youtube.com/watch?v=joQA2rb9Kdo
• https://aws.amazon.com/blogs/big-data/top-10-performance-tuning-tips-for-amazon-athena/