# Three data centers In Two cities (B)

+ Features:
  + Service availability is guaranteed without data loss if any data center fails
+ Question:
  + When two centers fail, most replica are not available for offsite disaster recovery, and the service is unavailable
  + Offsite disaster recover is asynchronous and does not guarantee consistency recovery
  + The high cost of dedicated network lines
------------------------------------------------------------
Three central structures in the same city

Advantages:
The distance between 2 cities is more than 1000 km. (Network latency is around 20 ms)
The same city network has easy latency (less than 1.5 ms) and bandwidth of 10 megabits.
When either hub fails, the service is available without loss of data.

Requirements:
The leader in TiKV in the disaster recovery city was removed due to unbearable network delays in different cities.
TiDB does not connect to disaster recovery cities.
The leader of PD is not in the disaster recovery center.
If the data centers of 2 centers in the same city are not available, there is a risk of unavailability or data loss.
The cost is extremely high.
Replication between the 2 hubs is asynchronous. If the primary hub fails, the disaster recovery center RPO may not be 0, but consistency is not guaranteed, for example, 2 regions, one updated and the other not, resulting in inconsistency.

To summarize:
Only room-level failures can be tolerated. After a primary city failure, the disaster city may lose data (RPO is not 0), and the recover data is not guaranteed to be consistent.