# Coprocessor

![left](./student-guide/diagram/03-Coprocessor-01-01.png)

------------------------------------------------------------
So far, our TiKV has solved the problem of data storage, so the entire TiDB database can be processed as shown below. TiDB is stateless, and the data is stored in the TiKV layer. When TiDB receives a query request from the client, it obtains specific data information from TiKV. There are 2 performance issues with this architecture:

  - TiKV return all data, which is too overhead the network.
  - TiDB needs to calculate all the data, and the CPU consumption is huge. Relatively, TiKV does not have much calculation, and the CPU pressure is very low