Comment by rbranson

You are correct. As a point of comparison: almost ten years ago at Segment we had a single Aurora PostgreSQL instance with ~50T of data, it was used to index potential identity data in a much larger corpus of files stored in S3.