# Three data centers within the Same City (C)


![left](./student-guide/diagram/29-Three-center-structure-in-the-same-city-03-01.png)

------------------------------------------------------------


Optimized:
If each data center does not need to provide services to the outside world at the same time, you can distribute all traffic to one data center, and migrate both the region leader and PD leader to the same data center through scheduling policies. This way, whether you get the TSO from the PD or read the region, it won't be affected by the inter-data center network. When the data center fails, the PD leader and region leader are automatically selected in the other data centers, and only need to shift the traffic to other surviving data centers.

Problems after optimization:
- Write scenario are still affected by data center network latency because following the Raft majority protocol, all written data needs to be replicated synchronously to at least 2 data centers
- TiDB Server data center level single point
- Business traffic is a single data center, performance is limited by the network bandwidth pressure of a single data center
- TSO acquisition and read performance are limited by the traffic data center cluster PD, TiKV components are normal, otherwise it is still affected by cross-data center network interaction