*创建未分区表
mysql> create table no_part_tab (c1 int(11) default NULL,c2 varchar(30) default NULL,c3 date default NULL) engine=myisam;
*通过存储过程灌入800万条测试数据
DELIMITER //
mysql> CREATE PROCEDURE load_part_tab()
begin
declare v int default 0;
while v < 8000000
do
insert into part_tab
values (v,'testing partitions',adddate('1995-01-01',(rand(v)*36520) mod 3652));
set v = v + 1;
end while;
end;
//
Query OK, 1 row affected (8 min 17.75 sec)
mysql> insert into no_part_tab select * from part_tab;
Query OK, 8000000 rows affected (51.59 sec)
Records: 8000000 Duplicates: 0 Warnings: 0
* 测试SQL性能
mysql> select count(*) from part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31';
+----------+
| count(*) |
+----------+
| 795181 |
+----------+
1 row in set (0.55 sec)
mysql> select count(*) from no_part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31';
+----------+
| count(*) |
+----------+
| 795181 |
+----------+
1 row in set (4.69 sec)
结果表明分区表比未分区表的执行时间少90%。
* 通过explain语句来分析执行情况
mysql > explain select count(*) from no_part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31'\G
/* 结尾的\G使得mysql的输出改为列模式 */
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: no_part_tab
type: ALL
possible_keys: NULL
key: NULL
key_len: NULL
ref: NULL
rows: 8000000
Extra: Using where
1 row in set (0.00 sec)
mysql> explain select count(*) from part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31'\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: part_tab
type: ALL
possible_keys: NULL
key: NULL
key_len: NULL
ref: NULL
rows: 798458
Extra: Using where
1 row in set (0.00 sec)
explain语句显示了SQL查询要处理的记录数目
* 试验创建索引后情况
mysql> create index idx_of_c3 on no_part_tab (c3);
Query OK, 8000000 rows affected (1 min 18.08 sec)
Records: 8000000 Duplicates: 0 Warnings: 0
mysql> create index idx_of_c3 on part_tab (c3);
Query OK, 8000000 rows affected (1 min 19.19 sec)
Records: 8000000 Duplicates: 0 Warnings: 0
创建索引后的数据库文件大小列表:
2008-05-24 09:23 8,608 no_part_tab.frm
2008-05-24 09:24 255,999,996 no_part_tab.MYD
2008-05-24 09:24 81,611,776 no_part_tab.MYI
2008-05-24 09:25 0 part_tab#P#p0.MYD
2008-05-24 09:26 1,024 part_tab#P#p0.MYI
2008-05-24 09:26 25,550,656 part_tab#P#p1.MYD
2008-05-24 09:26 8,148,992 part_tab#P#p1.MYI
2008-05-24 09:26 25,620,192 part_tab#P#p10.MYD
2008-05-24 09:26 8,170,496 part_tab#P#p10.MYI
2008-05-24 09:25 0 part_tab#P#p11.MYD
2008-05-24 09:26 1,024 part_tab#P#p11.MYI
2008-05-24 09:26 25,656,512 part_tab#P#p2.MYD
2008-05-24 09:26 8,181,760 part_tab#P#p2.MYI
2008-05-24 09:26 25,586,880 part_tab#P#p3.MYD
2008-05-24 09:26 8,160,256 part_tab#P#p3.MYI
2008-05-24 09:26 25,585,696 part_tab#P#p4.MYD
2008-05-24 09:26 8,159,232 part_tab#P#p4.MYI
2008-05-24 09:26 25,585,216 part_tab#P#p5.MYD
2008-05-24 09:26 8,159,232 part_tab#P#p5.MYI
2008-05-24 09:26 25,655,740 part_tab#P#p6.MYD
2008-05-24 09:26 8,181,760 part_tab#P#p6.MYI
2008-05-24 09:26 25,586,528 part_tab#P#p7.MYD
2008-05-24 09:26 8,160,256 part_tab#P#p7.MYI
2008-05-24 09:26 25,586,752 part_tab#P#p8.MYD
2008-05-24 09:26 8,160,256 part_tab#P#p8.MYI
2008-05-24 09:26 25,585,824 part_tab#P#p9.MYD
2008-05-24 09:26 8,159,232 part_tab#P#p9.MYI
2008-05-24 09:25 8,608 part_tab.frm
2008-05-24 09:25 68 part_tab.par
* 再次测试SQL性能
mysql> select count(*) from no_part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31';
+----------+
| count(*) |
+----------+
| 795181 |
+----------+
1 row in set (2.42 sec) /* 为原来4.69 sec 的51%*/
重启mysql ( net stop mysql, net start mysql)后,查询时间降为0.89 sec,几乎与分区表相同。
mysql> select count(*) from part_tab where c3 > date '1995-01-01' and c3 < date '1995-12-31';
+----------+
| count(*) |
+----------+
| 795181 |
+----------+
1 row in set (0.86 sec)
* 更进一步的试验
** 增加日期范围
mysql> select count(*) from no_part_tab where c3 > date '1995-01-01' and c3 < date '1997-12-31';
+----------+
| count(*) |
+----------+
| 2396524 |
+----------+
1 row in set (5.42 sec)
mysql> select count(*) from part_tab where c3 > date '1995-01-01' and c3 < date '1997-12-31';
+----------+
| count(*) |
+----------+
| 2396524 |
+----------+
1 row in set (2.63 sec)
** 增加未索引字段查询
mysql> select count(*) from part_tab where c3 > date '1995-01-01' and c3 < date
'1996-12-31' and c2='hello';
+----------+
| count(*) |
+----------+
| 0 |
+----------+
1 row in set (0.75 sec)
mysql> select count(*) from no_part_tab where c3 > date '1995-01-01' and c3 < date '1996-12-31' and c2='hello';
+----------+
| count(*) |
+----------+
| 0 |
+----------+
1 row in set (11.52 sec)
删除老数据,分区表的速度更快:
MySQL [Mytestdb]> alter table part_tab drop PARTITION p1;
速度比delete删除快很多;
= 初步结论 =
* 分区和未分区占用文件空间大致相同 (数据和索引文件)
* 如果查询语句中有未建立索引字段,分区时间远远优于未分区时间
* 如果查询语句中字段建立了索引,分区和未分区的差别缩小,分区略优于未分区。
= 最终结论 =
* 对于大数据量,建议使用分区功能。
* 去除不必要的字段
* 根据手册, 增加myisam_max_sort_file_size 会增加分区性能
***************************************** 其他类型 *********************************************
[分区命令详解]
= 分区例子 =
* RANGE 类型
CREATE TABLE users (
uid INT UNSIGNED NOT NULL AUTO_INCREMENT PRIMARY KEY,
name VARCHAR(30) NOT NULL DEFAULT '',
email VARCHAR(30) NOT NULL DEFAULT ''
)
PARTITION BY RANGE (uid) (
PARTITION p0 VALUES LESS THAN (3000000)
DATA DIRECTORY = '/data0/data'
INDEX DIRECTORY = '/data1/idx',
PARTITION p1 VALUES LESS THAN (6000000)
DATA DIRECTORY = '/data2/data'
INDEX DIRECTORY = '/data3/idx',
PARTITION p2 VALUES LESS THAN (9000000)
DATA DIRECTORY = '/data4/data'
INDEX DIRECTORY = '/data5/idx',
PARTITION p3 VALUES LESS THAN MAXVALUE DATA DIRECTORY = '/data6/data'
INDEX DIRECTORY = '/data7/idx'
);
在这里,将用户表分成4个分区,以每300万条记录为界限,每个分区都有自己独立的数据、索引文件的存放目录,与此同时,这些目录所在的物理磁盘分区可能也都是完全独立的,可以提高磁盘IO吞吐量。