Reviewed by Aditya Kumar · Last reviewed 2026-03-24
Architectural logic: Athena is serverless Presto/Trino—no cluster to manage; you pay per TB scanned. Why this model: Perfect for ad-hoc analytics on data lakes; decouples storage (S3) from compute. S3 interaction: Athena reads directly from S3; no ETL load step. Glue Catalog...
Red Flag: Queries without partition predicates on large tables—cost explosion. Pro-Move: 'We partition by date and use Parquet; our 100TB table costs $50/query instead of $500—we enforce partition filters in IAM.'
Architectural logic: Athena is serverless Presto/Trino—no cluster to manage; you pay per TB scanned. Why this model: Perfect for ad-hoc analytics on data lakes; decouples storage (S3) from compute. S3 interaction: Athena reads directly from S3; no ETL load step. Glue Catalog provides schema; queries scan only relevant partitions and columnar bytes. Scalability: Unlimited concurrency; each query spins up its own cluster. Cost implications: $5/TB scanned—partitioning and Parquet can reduce scans 10–100x vs CSV/JSON. Example: s3://bucket/events/year=2024/month=03/*.parquet—query with WHERE year=2024 AND month=3 prunes to March only. Trade-off: Cold start ~1–2s; for sub-second latency use Redshift or caching.
Red Flag: Queries without partition predicates on large tables—cost explosion. Pro-Move: 'We partition by date and use Parquet; our 100TB table costs $50/query instead of $500—we enforce partition filters in IAM.'
According to DataEngPrep.tech, this is one of the most frequently asked Cloud/Tools interview questions, reported at 1 company. DataEngPrep.tech maintains an editor-reviewed database of 1,863 data engineering interview questions across 7 categories.