A monitoring system writes 1 TB of metric records every minute to Amazon S3. Analysts query this data with Amazon Athena, but queries are slow because of the data volume. Which file format stored in S3 will improve Athena query performance?

Parquet files are a columnar storage format, which significantly improves Athena query performance. Because Athena can read only the necessary columns for a query, it reduces the amount of data scanned, leading to faster query execution and lower costs. CSV and JSON are row-based formats, meaning Athena must scan entire rows even if only a few columns are needed, making them less efficient for analytical queries on large datasets. While compression helps, the fundamental row-based nature remains a bottleneck. RecordIO is a record-oriented format often used in deep learning, but it doesn't offer the same columnar query optimization benefits as Parquet for Athena.