# Self-Increasing Id (Continue...)

+ `AUTO_INCREMENT` The principle implement
  + Each self-incrementing column uses a globally visible key-value pair to record the highest ID currently assigned
  + To reduce the network overhead of distributing self-incrementing IDs in distributed systems, each TiDB node caches a distinct ID segment
  + If the currently pre-allocated ID segment is used up, or TiDB is restart, a new ID segment will be requested again


![left](./student-guide/diagram/13-increment-ID-02-06.png)
------------------------------------------------------------
[no notes]