谨慎使用MYSQL表分区

test1表是未分区

test2表是hash分区

以下是两个表的表结构：

mysql> show create table test1;

+-------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

| Table | Create Table                                                                                                                                                                                      |

| test1 | CREATE TABLE `test1` (

  `id` int(11) NOT NULL,

  `tid` int(11) DEFAULT NULL,

  `name` varchar(10) DEFAULT NULL,

  PRIMARY KEY (`id`),

  KEY `tid` (`tid`)

) ENGINE=InnoDB DEFAULT CHARSET=latin1 |

1 row in set (0.00 sec)

mysql> show create table test2;

+-------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

| Table | Create Table                                                                                                                                                                                                                                       |

| test2 | CREATE TABLE `test2` (

) ENGINE=InnoDB DEFAULT CHARSET=latin1

/*!50100 PARTITION BY HASH (id)

PARTITIONS 10 */ |

1 row in set (0.01 sec)

在主键id列上用上hash分区。

如果是在主键上查询，那么分区要比不分区性能高，例如：

mysql> explain partitions select * from test1 where id in (86,777,432345);

+----+-------------+-------+------------+-------+---------------+---------+---------+------+------+-------------+

| id | select_type | table | partitions | type  | possible_keys | key     | key_len | ref  | rows | Extra       |

|  1 | SIMPLE      | test1 | NULL       | range | PRIMARY       | PRIMARY | 4       | NULL |    3 | Using where |

1 row in set (0.06 sec)

mysql> explain partitions select * from test2 where id in (86,777,432345); 

|  1 | SIMPLE      | test2 | p5,p6,p7   | range | PRIMARY       | PRIMARY | 4       | NULL |    3 | Using where |

只在分区P5/6/7上就扫描到了记录，减少了过多的磁盘IO。

但如果在索引tid上查询，性能就很低了。

mysql>  explain partitions select * from test1 where tid = 86419804; 

+----+-------------+-------+------------+------+---------------+------+---------+-------+------+-------------+

| id | select_type | table | partitions | type | possible_keys | key  | key_len | ref   | rows | Extra       |

|  1 | SIMPLE      | test1 | NULL       | ref  | tid           | tid  | 5       | const |    1 | Using where |

mysql>  explain partitions select * from test2 where tid = 86419804; 

+----+-------------+-------+-------------------------------+------+---------------+------+---------+-------+------+-------------+

| id | select_type | table | partitions                    | type | possible_keys | key  | key_len | ref   | rows | Extra       |

|  1 | SIMPLE      | test2 | p0,p1,p2,p3,p4,p5,p6,p7,p8,p9 | ref  | tid           | tid  | 5       | const |   10 | Using where |

要扫描10个分区，才能得到结果。

mysql> select sql_no_cache * from test1 where tid=86419804; 

+----+----------+------+

| id | tid      | name |

| 11 | 86419804 | abc2 |

1 row in set (0.40 sec)

mysql> select sql_no_cache * from test2 where tid=86419804; 

1 row in set (1.98 sec)

分区后的查询结果要比未分区的慢很多。

结论：分区规则仅限制在主键上使用，性能会有提高，在其他字段上使用，即使有索引，性能也会很差。

本文转自 liang3391 51CTO博客，原文链接:http://blog.51cto.com/liang3391/844523