A fleet of IoT devices writes data as ORC files into HDFS on a persistent Amazon EMR cluster. Analysts run SQL queries with Apache Presto on the same EMR cluster. Queries scan large volumes, always complete in under 15 minutes, and run only between 5 PM and 10 PM. The company wants a more cost-effective solution that still allows SQL querying. Which solution is the most cost-effective and meets the requirements?

Storing data in Amazon S3 and querying with Amazon Athena is the most cost-effective solution. Athena is serverless, meaning you only pay for the data scanned, which aligns well with the defined query window (5 PM - 10 PM) and the need for cost savings. The AWS Glue Data Catalog provides the necessary metadata for Athena to understand the ORC files. Storing data in Amazon S3 and using Amazon Redshift Spectrum is less cost-effective because Redshift Spectrum still requires a running Redshift cluster, incurring compute costs even when not querying. Keeping data on a persistent EMR cluster is expensive as you pay for the cluster 24/7, even outside the query window. Loading data into Amazon Redshift is also costly due to continuous cluster charges and the overhead of data ingestion, which is unnecessary given the ORC file format and existing query patterns.