Mysql database and table splitting strategy and application scenarios

sub-treasury

Solve the performance problem of a single database

Horizontal sub database ： modulus

Determine how many libraries the data needs to be evenly distributed , That is to say ： Storage capacity

For users ID Take the mold ,user_id% Storage capacity , The result is 0 In the first library , The result is 1 In the second library ...

Vertical sub database ： Split according to business

When there is too much order data, a separate database can be created for the order

table

Solve the query performance problem of single table massive data

Horizontal sub table

Split data in behavioral units

After the sub table , The structure of all tables is the same

Solve the problem of too much table data

Method ： Determine how many tables the data needs to be evenly distributed , That is to say ： Table capacity

            For users ID Take the mold ,user_id% Table capacity , The result is 0 In the first table , The result is 1 In the second table ...

Vertical sub table

Split into multiple tables by column , Multiple tables are associated through primary keys , So as to form a complete data

After the sub table , The structure of each table is different

Solve the problem of too many columns

Simultaneous database and table

Intermediate variable 　＝ user_id%（ Storage capacity * Table capacity per library ）

Library number 　＝　 integer （ Intermediate variable ／ Table capacity per library ）

Table number 　＝　 Intermediate variable ％ Table capacity per library

  disadvantages

Originally, things across tables have become distributed things ;

Because records are segmented into different databases and different data tables , It is difficult to perform multi table Association query , And query the library number and table number of the data to be specified .

Further expand the system （ Change of warehouse number and table number ）, Will become very inconvenient , We need to re migrate data .

How much data scale is appropriate

theory 500w Separable table （ It is faster after the table is divided ）

actual 1000w-2000w Left and right can be divided into two parts

In short, it should not exceed 5000w Talent score table ,5000w Is the index limit

Add ： If the order data needs to be counted , Statistics of monthly sales and other information , You can use a monthly schedule , otherwise , Don't suggest , Because there may be too many data in a certain month , However, there are few data in a certain month