代码改变世界

MySQL 5.6新特性 -- Index Condition Pushdown

  abce  阅读(1568)  评论(0)  编辑  收藏  举报
Index Condition Pushdown(ICP)是针对mysql使用索引从表中检索行数据时的一种优化方法。
 
在没有ICP特性之前，存储引擎根据索引去基表查找并将数据返回给mysql server，mysql server再根据where条件进行数据过滤。
有了ICP之后，在取出索引的同时，判断是否可以根据索引中的列进行where条件过滤，也就是将where的部分过滤操作放在了存储引擎层。这样就会减少上层sql层对记录的获取。
 
ICP优化支持range、ref、eq_ref、ref_or_null类型的查询。查询优化器会给出相应的提示：Using index condition。当开启ICP后，在执行计划的extra列会显示：Using index condition。
 
ICP支持innodb、myisam表。对于innodb表，ICP只是用于辅助索引。
 
在5.6中，ICP不支持分区表。这个问题在mysql 5.7中得到解决。
 
优化器使用ICP时，server层将会把能够通过使用索引进行评估的where条件下推到storage engine层。数据访问和提取过程如下：
1) storage engine从索引中读取下一条索引元组。
2) storage engine使用索引元组评估下推的索引条件。如果没有满足where条件，storage engine将会处理下一条索引元组（回到上一步）。只有当索引元组满足下推的索引条件的时候，才会继续去基表中读取数据。
3) 如果满足下推的索引条件，storage engine通过索引元组定位基表的行和读取整行数据并返回给server层。
4) server层评估没有被下推到storage engine层的where条件，如果该行数据满足where条件则使用，否则丢弃。
 
没有ICP之前：
 
开启ICP之后，就变成：
默认是开启ICP的，手动开启/关闭ICP： 
?
1
2
set optimizer_switch = 'index_condition_pushdown=off';
set optimizer_switch = 'index_condition_pushdown=on';

　　

 

 测试过程

 1.环境准备
?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
#mysql 5.6.25
#关闭结果缓存
mysql> set global query_cache_size=0;
mysql> set query_cache_type=off;
 
#查看表结构
mysql> show create table employees\G
*************************** 1. row ***************************
       Table: employees
Create Table: CREATE TABLE `employees` (
  `emp_no` int(11) NOT NULL,
  `birth_date` date NOT NULL,
  `first_name` varchar(14) NOT NULL,
  `last_name` varchar(16) NOT NULL,
  `gender` enum('M','F') NOT NULL,
  `hire_date` date NOT NULL,
  PRIMARY KEY (`emp_no`),
  KEY `idx_first_last_name` (`first_name`,`last_name`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8
1 row in set (0.00 sec)
 
mysql>

2.开启ICP后进行测试

?
1
2
3
4
5
mysql> set profiling = 1;
mysql> select * from employees where first_name='Anneke' and last_name like '%sig' ;
mysql> explain select * from employees where first_name='Anneke' and last_name like '%sig' ;
mysql> show profiles;
mysql> show profile cpu,block io for query 1;

3.关闭ICP后进行测试

?
1
2
3
4
5
6
mysql> set optimizer_switch='index_condition_pushdown=off';
mysql> set profiling = 1;
mysql> select * from employees where first_name='Anneke' and last_name like '%sig' ;
mysql> explain select * from employees where first_name='Anneke' and last_name like '%sig' ;
mysql> show profiles;
mysql> show profile cpu,block io for query 1;

4.结果比较

开启ICP后的执行计划：执行计划中extra部分的内容是"using index condition"

?
1
2
3
4
5
6
mysql> explain select * from employees where first_name='Anneke' and last_name like '%sig' ;
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-----------------------+
| id | select_type | table     | type | possible_keys       | key                 | key_len | ref   | rows | Extra                 |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-----------------------+
|  1 | SIMPLE      | employees | ref  | idx_first_last_name | idx_first_last_name | 44      | const |  224 | Using index condition |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-----------------------+

关闭ICP后的执行计划：执行计划中extra部分的内容是"using where"

?
1
2
3
4
5
6
mysql> explain select * from employees where first_name='Anneke' and last_name like '%sig' ;
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-------------+
| id | select_type | table     | type | possible_keys       | key                 | key_len | ref   | rows | Extra       |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-------------+
|  1 | SIMPLE      | employees | ref  | idx_first_last_name | idx_first_last_name | 44      | const |  224 | Using where |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------+------+-------------+

　

开启ICP后的profile内容：Sending data部分的值是0.000212s

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
mysql> show profile cpu,block io for query 1;
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000114 | 0.000000 |   0.000000 |            0 |             0 |
| checking permissions | 0.000007 | 0.000000 |   0.000000 |            0 |             0 |
| Opening tables       | 0.000018 | 0.000000 |   0.000000 |            0 |             0 |
| init                 | 0.000034 | 0.000000 |   0.000000 |            0 |             0 |
| System lock          | 0.000008 | 0.000000 |   0.000000 |            0 |             0 |
| optimizing           | 0.000023 | 0.000000 |   0.000000 |            0 |             0 |
| statistics           | 0.000383 | 0.000000 |   0.000000 |            0 |             0 |
| preparing            | 0.000019 | 0.000000 |   0.000000 |            0 |             0 |
| executing            | 0.000002 | 0.000000 |   0.000000 |            0 |             0 |
| Sending data         | 0.000212 | 0.000000 |   0.000000 |            0 |             0 |
| end                  | 0.000004 | 0.000000 |   0.000000 |            0 |             0 |
| query end            | 0.000004 | 0.000000 |   0.000000 |            0 |             0 |
| closing tables       | 0.000006 | 0.000000 |   0.000000 |            0 |             0 |
| freeing items        | 0.000020 | 0.000000 |   0.000000 |            0 |             0 |
| cleaning up          | 0.000011 | 0.000000 |   0.000000 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+

关闭ICP后的profile内容：Sending data部分的值是0.010990s

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
mysql> show profile cpu,block io for query 1;
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000165 | 0.000000 |   0.000000 |            0 |             0 |
| checking permissions | 0.000022 | 0.000000 |   0.000000 |            0 |             0 |
| Opening tables       | 0.000027 | 0.000000 |   0.000000 |            0 |             0 |
| init                 | 0.000039 | 0.000000 |   0.000000 |            0 |             0 |
| System lock          | 0.000008 | 0.000000 |   0.000000 |            0 |             0 |
| optimizing           | 0.000037 | 0.001000 |   0.000000 |            0 |             0 |
| statistics           | 0.000483 | 0.001000 |   0.000000 |            0 |             0 |
| preparing            | 0.000022 | 0.000000 |   0.000000 |            0 |             0 |
| executing            | 0.000002 | 0.000000 |   0.000000 |            0 |             0 |
| Sending data         | 0.010990 | 0.007999 |   0.002000 |            0 |             0 |
| end                  | 0.000009 | 0.000000 |   0.000000 |            0 |             0 |
| query end            | 0.000005 | 0.000000 |   0.000000 |            0 |             0 |
| closing tables       | 0.000008 | 0.000000 |   0.000000 |            0 |             0 |
| freeing items        | 0.000028 | 0.000000 |   0.000000 |            0 |             0 |
| cleaning up          | 0.000014 | 0.000000 |   0.000000 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+

　　

其它：

当sql使用覆盖索引时，不支持ICP优化方法

?
1
2
3
4
5
6
7
8
9
10
11
12
mysql> explain select first_name,last_name from employees where first_name='Anneke' and last_name='Porenta' ;
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+--------------------------+
| id | select_type | table     | type | possible_keys       | key                 | key_len | ref         | rows | Extra                    |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+--------------------------+
|  1 | SIMPLE      | employees | ref  | idx_first_last_name | idx_first_last_name | 94      | const,const |    1 | Using where; Using index |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+--------------------------+
mysql> explain select * from employees where first_name='Anneke' and last_name='Porenta' ;
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+-----------------------+
| id | select_type | table     | type | possible_keys       | key                 | key_len | ref         | rows | Extra                 |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+-----------------------+
|  1 | SIMPLE      | employees | ref  | idx_first_last_name | idx_first_last_name | 94      | const,const |    1 | Using index condition |
+----+-------------+-----------+------+---------------------+---------------------+---------+-------------+------+-----------------------+

　　

编辑推荐：
· AI与.NET技术实操系列（二）：开始使用ML.NET
· 记一次.NET内存居高不下排查解决与启示
· 探究高空视频全景AR技术的实现原理
· 理解Rust引用及其生命周期标识（上）
· 浏览器原生「磁吸」效果！Anchor Positioning 锚点定位神器解析
阅读排行：
· DeepSeek 开源周回顾「GitHub 热点速览」
· 物流快递公司核心技术能力-地址解析分单基础技术分享
· .NET 10首个预览版发布：重大改进与新特性概览！
· AI与.NET技术实操系列（二）：开始使用ML.NET
· .NET10 - 预览版1新功能体验（一）
历史上的今天：
2015-11-06 向数据库中导入AWR数据
2015-11-06 抽取AWR数据
点击右上角即可分享
微信分享提示
AI FOR CODE 大赛