mysql前缀索引区分度长度计算 mysql前缀索引 实现

接近三年的MySQL 运维经验，近期决定把曾经做过的优化案例拿出来和大家分享！

MySQL优化绝对要了解业务，和开发沟通才能做到最好。以下是一个小案例，说明一下前缀索引的用法。

1.查看表结构
(root@localhost) [prod_db]> show create table t_file_info\G;
*************************** 1. row ***************************
       Table: t_file_info
Create Table: CREATE TABLE `t_file_info` (
  `id` varchar(36) NOT NULL DEFAULT '',
  `devid` varchar(64) DEFAULT NULL,
  `areaid` int(11) DEFAULT NULL,
  `fileid` varchar(256) NOT NULL,
  `filename` varchar(256) DEFAULT NULL,
  `filesize` int(11) DEFAULT NULL,
  `filemd5` varchar(40) DEFAULT NULL,
  `extend` varchar(4000) DEFAULT NULL,
  `status` int(11) NOT NULL DEFAULT '0',
  `createdate` datetime DEFAULT NULL,
  `fileurl` varchar(256) DEFAULT NULL,
  `businessid` bigint(20) NOT NULL DEFAULT '0',
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8

2.fileid是我们查询的一个条件，正常是需要创建索引的。

select char_length('63f2a078018649ca9948f5469550bf2a/group1/M00/00/DA/wKgj2FcMquGAVuJcAAAI4FL7ZCA388.jpg') ;
+-----------------------------------------------------------------------------------------------------+
| char_length('63f2a078018649ca9948f5469550bf2a/group1/M00/00/DA/wKgj2FcMquGAVuJcAAAI4FL7ZCA388.jpg') |
+-----------------------------------------------------------------------------------------------------+
|                                                                                                  84 |
+-----------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)
--经过和开发沟通了解，前32位相当于uuid可以确定唯一值。

3.这样的字段，我们怎么创建索引，是不是有规律可循。继续查看

--查看选择率
select count(distinct(fileid))/count(*) AS Selectivity from t_file_info;

select count(distinct left(fileid,32))/count(*) from t_file_info;

(root@localhost) [prod_db]> select count(distinct(fileid))/count(*) from t_file_info;
+----------------------------------+
| count(distinct(fileid))/count(*) |
+----------------------------------+
|                           1.0000 |
+----------------------------------+
1 row in set (0.17 sec)

(root@localhost) [prod_db]> select count(distinct left(fileid,32))/count(*) from t_file_info;
+------------------------------------------+
| count(distinct left(fileid,32))/count(*) |
+------------------------------------------+
|                                   0.9999 |
+------------------------------------------+
1和0.9999几乎可以等同，其实这里因为点特殊情况，正常应该都是1才对的。

4.查看无索引的执行计划

explain select id,fileid from prod_db.t_file_info where fileid='63f2a078018649ca9948f5469550bf2a/group1/M00/00/DA/wKgj2FcMquGAVuJcAAAI4FL7ZCA388.jpg';

(root@localhost) [prod_db]> explain select id,fileid from prod_db.t_file_info where fileid='63f2a078018649ca9948f5469550bf2a/group1/M00/00/DA/wKgj2FcMquGAVuJcAAAI4FL7ZCA388.jpg';
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
| id | select_type | table       | type | possible_keys | key  | key_len | ref  | rows  | Extra       |
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
|  1 | SIMPLE      | t_file_info | ALL  | NULL          | NULL | NULL    | NULL | 35109 | Using where |
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
1 row in set (0.00 sec)

5.创建前缀索引，查看执行计划

alter table `prod_db`.`t_file_info` add index idx_t_file_info_fileid(fileid(32));

(root@localhost) [prod_db]> explain select id,fileid from prod_db.t_file_info where fileid='63f2a078018649ca9948f5469550bf2a/group1/M00/00/DA/wKgj2FcMquGAVuJcAAAI4FL7ZCA388.jpg';
+----+-------------+-------------+------+------------------------+------------------------+---------+-------+------+-------------+
| id | select_type | table       | type | possible_keys          | key                    | key_len | ref   | rows | Extra       |
+----+-------------+-------------+------+------------------------+------------------------+---------+-------+------+-------------+
|  1 | SIMPLE      | t_file_info | ref  | idx_t_file_info_fileid | idx_t_file_info_fileid | 98      | const |    1 | Using where |
+----+-------------+-------------+------+
--返回1行才是我们想看到的

6.创建索引

(root@localhost) [prod_db]> alter table `prod_db`.`t_file_info` add index idx_t_file_info_fileid(fileid(32));
Query OK, 0 rows affected (5 min 36.03 sec)
Records: 0  Duplicates: 0  Warnings: 0

创建索引观察系统资源使用情况，内存机会没有变化，但是CPU单核几乎跑满

(root@localhost) [prod_db]> select count(fileid) from t_file_info;
+---------------+
| count(fileid) |
+---------------+
|      12299419 |
+---------------+
1 row in set (14.94 sec) --千万行

小结：
1.了解前缀索引的实用场景。
2.要和开发沟通，了解业务，才能创建最合适的索引。
3.创建索引对系统性能会有很大的影响，要选择一个合适的时间点去创建，评估好影响。任何事情不要想当然，当你没经验，还想当然的时候很容易出问题。

http://blog.51cto.com/roidba/2051546