An online retailer stores ALB access logs in S3 and queries them with Amazon Athena. The engineer created an unpartitioned Athena table, and query response times have grown as the data volume increased. Which approach will improve Athena query performance with the least operational effort?

The correct answer is to create an AWS Glue crawler because it automates schema inference and partition discovery for S3 data, directly updating the AWS Glue Data Catalog which Athena uses. This significantly improves query performance by allowing Athena to scan only relevant data partitions, and it requires minimal operational effort as Glue handles the discovery process. Creating an AWS Glue job is less efficient because jobs are typically used for data transformation, not just metadata discovery, and inferring the schema for all logs without partitioning would still lead to full table scans. Creating an AWS Lambda function to transform logs to Parquet and add partitions is effective for performance but involves significant operational overhead for development and maintenance. Using Apache Hive for bucketed tables is also a valid performance improvement technique but introduces external dependencies and more operational complexity compared to a Glue crawler.