SQL优化：limit分页优化

分页查询

分页查询的问题点主要集中在

1. 如何快速定位起始点
2. 减少无用数据缓存

mysql为分页查询提供了很方便的关键字limit，但这个关键字在数据量较大时，却很低效。
“limit m,n”关键字的意思是，从第m行开始，扫描满足条件的n个偏移行。若需从第1行开始，则不需要指定m值。

示例

表aaaaa中共有2375690数据。

优化前的SQL

SQL执行结果：

SELECT DISTINCT(device_id) uid FROM aaaaa WHERE status = 0 LIMIT 88000,1000;
1000 rows in set (0.48 sec)

SQL执行计划：

MariaDB [star]> explain SELECT sql_no_cache DISTINCT(device_id) uid FROM aaaaa WHERE status = 0 LIMIT 88000,1000;
+------+-------------+---------------+------+---------------+------+---------+------+---------+------------------------------+
| id   | select_type | table         | type | possible_keys | key  | key_len | ref  | rows    | Extra                        |
+------+-------------+---------------+------+---------------+------+---------+------+---------+------------------------------+
|    1 | SIMPLE      | aaaaa         | ALL  | NULL          | NULL | NULL    | NULL | 2375690 | Using where; Using temporary |
+------+-------------+---------------+------+---------------+------+---------+------+---------+------------------------------+

优化方式

迅速定位起始ID，利用主键索引，加快扫描速度。可以看到，derived中，SQL使用到了覆盖索引进行扫描，虽然还是全表扫，因为只扫描id列，大大降低了扫描的IO耗费，快速定位到了id。

MariaDB [star]> explain SELECT sql_no_cache DISTINCT(device_id) uid FROM aaaaa join (select id from aaaaa limit 88000,1) k on star_device_5.id>=k.id where status=0 limit 1000;
+------+-------------+---------------+-------+---------------+-------------+---------+------+---------+------------------------------------------------+
| id   | select_type | table         | type  | possible_keys | key         | key_len | ref  | rows    | Extra                                          |
+------+-------------+---------------+-------+---------------+-------------+---------+------+---------+------------------------------------------------+
|    1 | PRIMARY     |     | ALL   | NULL          | NULL        | NULL    | NULL |   88001 | Using temporary                                |
|    1 | PRIMARY     | star_device_5 | ALL   | PRIMARY       | NULL        | NULL    | NULL | 2377112 | Range checked for each record (index map: 0x1) |
|    2 | DERIVED     | star_device_5 | index | NULL          | idx_star_id | 8       | NULL | 2377112 | Using index                                    |
+------+-------------+---------------+-------+---------------+-------------+---------+------+---------+------------------------------------------------+

执行结果：

SELECT sql_no_cache DISTINCT(device_id) uid FROM star_device_5 join (select id from star_device_5 limit 880000,1) k on star_device_5.id>=k.id where status=0 limit 1000;
1000 rows in set (0.19 sec)

随着m的增大和n的增大，两种写法的SQL执行时间会有本质差别。我做了测试，当m值增加到880000时，优化前的SQL需要2分钟，优化后的SQL还是0.1s左右。