mysql - 加快 MYSQL 具有 order by 的查询速度

Question

我在“tblmessages”中有 300,000 行，并且我正在尝试运行此查询。 如果你不使用“order by msgId desc”它运行得非常快，但是当我添加订单时它非常非常慢。

我错过了什么......？

my index

    SELECT msgId  
    FROM tblmessages 
    left join  wp_users as a on a.ID  = (CASE WHEN (msgFromUserId=1) then tblmessages.msgToUserId else  tblmessages.msgFromUserId END)
    left join tblforum_users u1 on u1.user_ID  =(CASE WHEN (msgFromUserId=1) then tblmessages.msgToUserId else  tblmessages.msgFromUserId END)
    where (msgFromUserId=1 or msgToUserId=1) 
    order by msgId desc limit 0,20 

解释后:

+----+-------------+-------------+------------+-------------+---------------------------------------+---------------------------------------+---------+------+--------+----------+-----------------------------------------------------+
| id | select_type | table       | partitions | type        | possible_keys                         | key                                   | key_len | ref  | rows   | filtered | Extra                                               |
+----+-------------+-------------+------------+-------------+---------------------------------------+---------------------------------------+---------+------+--------+----------+-----------------------------------------------------+
| 1  | SIMPLE      | tblmessages | NULL       | index_merge | IXtbl_messages_from,IX_tblmessages_to | IX_tblmessages_from,IX_tblmessages_to | 5,5     | NULL | 726454 | 100.00   | Using union(IX_tblmessages_from,IX_tblmessages_to); |
                                                                                                                                                                                   Using where; Using filesort                         |
| 1  | SIMPLE      | a           | NULL       | eq_ref      | PRIMARY                               | PRIMARY                               | 8       | func | 1      | 100.00   | Using where; Using index                            |
| 1  | SIMPLE      | u1          | NULL       | eq_ref      | PRIMARY                               | PRIMARY                               | 4       | func | 1      | 100.00   | Using where; Using index                            |
+----+-------------+-------------+------------+-------------+---------------------------------------+---------------------------------------+---------+------+--------+----------+-----------------------------------------------------+

以及有关tblmessages的更多信息:

Data    12.1    GiB
Index   190.8   MiB
Total   12.3    GiB

Answer 1

造成这种情况的原因有两个:

If U don't use "order by msgId desc" It's run very fast, but when I add the order It's very very slow.

第一个是您的查询末尾有一个“LIMIT 20”行。对此查询的结果进行排序(通过“order by”)将需要所有满足where子句的行被拉出/生成、排序等。为了得到正确的 20 行。

第二是有点主观，但我会说你的表没有必要的索引来支持该查询的优化。或者，从另一种角度来看，这样的查询通常会因为 IF-THEN 子句上的多个联接而运行缓慢。我对解决此问题的建议是将查询简化为多个查询，或者使用相关表的“SHOW CREATE TABLE”语句以及汇总索引、分区等的 INFORMATION_SCHEMA 查询之一来更新您的问题。从那里我们可能会更新设计(例如添加多列索引/索引/无论其拼写如何......)