Unique ID Generation

Distributed writers still need unique identities#

A single database sequence is wonderfully simple. When many regions or shards must generate IDs independently, the design must trade central coordination against size, ordering, index locality, clock behavior, and collision risk.

Snowflake-style IDs combine locally available fields#

┌──────────── timestamp ────────────┬── node ──┬─ sequence ─┐
│ rough time ordering              │ writer   │ same-tick  │
└──────────────────────────────────┴───────────┴────────────┘

The node field prevents writers from sharing the same sequence space; the sequence handles multiple IDs in one time bucket. Plan for clock rollback, duplicate node IDs, and sequence exhaustion. Exposed structure can leak approximate time or traffic information.