一个问烂了的问题--数据量大的情况下分页查询很慢，有什么优化方案（各blog集合）

我自己有一张表，有100w条数据

limit分页时间对比#

同一起始位置，不同的偏移量#

类似下图

select * from poison.poison_users limit 10000, 10;

468 ms

select * from poison.poison_users limit 10000, 100;

515 ms

select * from poison.poison_users limit 10000, 1000;

541 ms

select * from poison.poison_users limit 10000, 10000;

2772 ms

select * from poison.poison_users limit 10000, 100000;

84059 ms

同样偏移不同起始位置#

类似下图

select * from poison.poison_users limit 100, 100;

482 ms

select * from poison.poison_users limit 1000, 100;

533 ms

select * from poison.poison_users limit 100000, 100;

555 ms

select * from poison.poison_users limit 1000000, 100;

3568 ms

limit 初始位置和偏移量都对查询速度都有影响

在无过滤条件下#

分库分表，采用独立的中间件 mycat 或者 sharding-jdbc 或者全套的ShardingSphere

在有过滤条件下#

使用子查询优化#

不用子查询的情况

select * from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' limit 100000,100;
select * from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' limit 100000,1;

1261 ms

select * from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' and
        id>=(select id from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' limit 100000,1)
limit 100;

470 ms

限定id范围#

select * from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' and id between 1000000 and 1000100;

490 ms

子查询中使用limit#

由于一些mysql一些版本中不支持子查询中limit，可以嵌套一层实现

select * from poison.poison_users a where id in (select u.id from 
(select id from poison.poison_users a where a.avatar='1231j4j1b4j4b1jk4b1k41b1j4' limit 100000,100) as u);