S3 performance is all about parallelization — doing multiple things at once instead of one at a time.
S3 is automatically fast out of the box with no configuration needed.
A prefix is simply the folder path inside your bucket.
bucket/folder1/sub1/ <- one prefix
bucket/folder1/sub2/ <- another prefix
bucket/photos/2024/ <- another prefix
Each prefix gets its own independent request limit.
Spread your files across multiple prefixes to multiply your throughput.
1 prefix = 5,500 GET requests/sec
4 prefixes = 22,000 GET requests/sec
10 prefixes = 55,000 GET requests/sec
Instead of putting all files in one folder, split them across several folders.