Best practices for designing S3 prefix for relational DB

[StrikeW]

Hello, I am curious about how Rockset designs object prefixes for those SST files on S3.

Amazon claims that each prefix can support at least 3,500 PUT/COPY/POST/DELETE or 5,500 GET/HEAD requests per second. If all SST files store in the root path of the bucket, e.g, s3:https://bucket/, then all requests will to go to the same prefix and meet the performance limits. Would you mind sharing your experience on this?