A new analytics platform stores large append-only datasets and runs periodic SQL analysis. The team wants minimal database administration and low storage cost. Which design is appropriate?

1. Use one unpartitioned text object and require every query to scan all records.
2. Load every historical byte into a single in-memory Redis node.
3. Store partitioned columnar data in S3 and query it with Athena using a catalog.
4. Use DynamoDB scans as the only SQL analytics engine for the entire archive.