Mysql5.7 partition table

Alibaba cloud mysql Basic information
edition ：5.7.32
To configure ：2 nucleus 4G
Data sources ：https://www.jianshu.com/p/d5d4fa4bd713
Table structure ：
Partition table structure

No partition table structure

Data volume :1 Billion

Refer to a Fei's blog , website ：https://www.jianshu.com/p/d5d4fa4bd713, A Fei's blog does not give a comparison of query performance before and after partitioning , Self test .
The following is the test primary key index
Test partition table sql

select * from employees_partition where job_no = ‘NO.8888891’;

Five times in a row ：
0.288s
1.552s
1.367s
0.444s
0.461s

Test partition off cache sql

select SQL_NO_CACHE * from employees_partition where job_no =
‘NO.8888891’;

Five times in a row ：
0.290s
0.438s
0.566s
0.533s
0.908s

Tests do not partition sql

select * from employees_nopartition where job_no = ‘NO.8888891’;

Five times in a row ：
0.355s
0.440s
0.523s
0.326s
0.567s

Test closing cache without partitioning sql

select SQL_NO_CACHE * from employees_nopartition where job_no =
‘NO.8888891’;

Five times in a row ：
0.477s
0.630s
0.325s
0.495s
0.285s

It is not found that partitioned tables have better performance than non partitioned queries .
If you are interested, you can self test the performance comparison of ordinary indexes .
Consider the worst case scenario of a large single table with large primary key fields , Let's calculate the main table b+ The height of the tree .
Like a single watch 100 Billion rows , Each row of data occupies an average of 1000 Byte storage space ,16KB Of page size, The main table page node will occupy about 10TB Space , about 7 Billion pages .
Suppose that the primary key index fetching is so stupid that each index row of the inner node occupies 512 byte , So the main table b+ The height of the tree is

1 + ceiling(lg(32, 700000000)) = 7

Assume that the above data only 1 Ten million lines , Then calculate the main table in the same way b+ The height of the tree is 5, Height difference 2.

Suppose the single table in the example 100 Yi did it 1000 Zones , Then through the partition , Can make each tree search reduce 2 Secondary page acquisition , This difference does lead to improved query performance , But the difference is not big .

On the other hand mysql5.7 In an earlier version , The implementation performance of partitioned table is poor , Compared with a single table with the same amount of data, the performance decreases by about 10%. Later on mysql5.7.19 To optimize , You can go to http://bugs.mysql.com Take a look at this bug. But even this bug After the repair , Partitioned tables are still about... Larger than single tables with the same amount of data 5% Performance degradation of （8 Zones ,100GB Data volume ,sysbench oltp Measurement example ）.

above , It is not recommended to use mysql Partition table .