Interesting. "KSUIDs are larger than UUIDs and Flake IDs, weighing in at 160 bits. They consist of a 32-bit timestamp and a 128-bit randomly generated payload.

Interesting. "KSUIDs are larger than UUIDs and Flake IDs, weighing in at 160 bits. They consist of a 32-bit timestamp and a 128-bit randomly generated payload." TTTTTTTTRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRRR T-Timestamp R-Random There are economy arguments against using this as quasi-sequential PKs. For efficient index lookup and sorting, best is to use CPU's native register sizes. It would be great if the PK would be a DWORD (32 bits) or QWORD (64 bits), one would assume the DB engine will align the access for efficient handling of the L1-L2 caches and prevent page tearing.