Thread (1): buckets for data. I assume you're asking about (S3) buckets in the underlying storage. I've typically seen very pragmatic rather than principled approaches. There are very few differences between buckets and prefixes within a bucket on S3. I think one important one is costs tooling - it's easier to separate costs between buckets than between prefixes. Even more pragmatically, someone has already made the decision for you: for instance, if I'm only allocated the one bucket name by my organization then for that scope of usage I have my answer. I do not think anyone has written this up anywhere. (And this could be a great place for people to share theirs, hint hint!)