[MySQL优化案例]系列 — 优化InnoDB表BLOB列的存储效率
时间:2016-05-29 12:23 来源:imysql.com 作者:imysql
首先，介绍下关于InnoDB引擎存储格式的几个要点：
1、InnoDB可以选择使用共享表空间或者是独立表空间方式，建议使用独立表空间，便于管理、维护。启用 innodb_file_per_table 选项，5.5以后可以在线动态修改生效，并且执行 ALTER TABLE xx ENGINE = InnoDB 将现有表转成独立表空间，早于5.5的版本，修改完这个选项后，需要重启才能生效；
2、InnoDB的data page默认16KB，5.6版本以后，新增选项 innodb_page_size 可以修改，在5.6以前的版本，只能修改源码重新编译，但并不推荐修改这个配置，除非你非常清楚它有什么优缺点；
3、InnoDB的data page在有新数据写入时，会预留1/16的空间，预留出来的空间可用于后续的新纪录写入，减少频繁的新增data page的开销；
4、每个data page，至少需要存储2行记录。因此理论上行记录最大长度为8KB，但事实上应该更小，因为还有一些InnoDB内部数据结构要存储；
5、受限于InnoDB存储方式，如果数据是顺序写入的话，最理想的情况下，data page的填充率是15/16，但一般没办法保证完全的顺序写入，因此，data page的填充率一般是1/2到15/16。因此每个InnoDB表都最好要有一个自增列作为主键，使得新纪录写入尽可能是顺序的；
6、当data page填充率不足1/2时，InnoDB会进行收缩，释放空闲空间；
7、MySQL 5.6版本的InnoDB引擎当前支持COMPACT、REDUNDANT、DYNAMIC、COMPRESSED四种格式，默认是COMPACT格式，COMPRESSED用的很少且不推荐（见下一条），如果需要用到压缩特性的话，可以直接考虑TokuDB引擎；
8、COMPACT行格式相比REDUNDANT，大概能节省20%的存储空间，COMPRESSED相比COMPACT大概能节省50%的存储空间，但会导致TPS下降了90%。因此强烈不推荐使用COMPRESSED行格式；
9、当行格式为DYNAMIC或COMPRESSED时，TEXT/BLOB之类的长列（long column，也有可能是其他较长的列，不一定只有TEXT/BLOB类型，看具体情况）会完全存储在一个独立的data page里，聚集索引页中只使用20字节的指针指向新的page，这就是所谓的off-page，类似ORACLE的行迁移，磁盘空间浪费较严重，且I/O性能也较差。因此，强烈不建议使用BLOB、TEXT、超过255长度的VARCHAR列类型；
10、当InnoDB的文件格式（innodb_file_format）设置为Antelope，并且行格式为COMPACT 或 REDUNDANT 时，BLOB、TEXT或者长VARCHAR列只会将其前768字节存储在聚集索页中（最大768字节的作用是便于创建前缀索引/prefix index），其余更多的内容存储在额外的page里，哪怕只是多了一个字节。因此，所有列长度越短越好；
11、在off-page中存储的BLOB、TEXT或者长VARCHAR列的page是独享的，不能共享。因此强烈不建议在一个表中使用多个长列。
综上，如果在实际业务中，确实需要在InnoDB表中存储BLOB、TEXT、长VARCHAR列时，有下面几点建议：
1、尽可能将所有数据序列化、压缩之后，存储在同一个列里，避免发生多次off-page；
2、实际最大存储长度低于255的列，转成VARCHAR或者CHAR类型（如果是变长数据二者没区别，如果是定长数据，则使用CHAR类型）；
3、如果无法将所有列整合到一个列，可以退而求其次，根据每个列最大长度进行排列组合后拆分成多个子表，尽量是的每个子表的总行长度小于8KB，减少发生off-page的频率；
4、上述建议是在data page为默认的16KB前提下，如果修改成8KB或者其他大小，请自行根据上述理论进行测试，找到最合适的值；
5、字符型列长度小于255时，无论采用CHAR还是VARCHAR来存储，或者把VARCHAR列长度定义为255，都不会导致实际表空间增大；
6、一般在游戏领域会用到比较多的BLOB列类型，游戏界同行可以关注下。
下面是测试验证过程，有耐心的同学可以慢慢看：
#
# 测试案例：InnoDB中长列存储效率
# 测试场景描述：
# 在InnoDB表中存储64KB的数据，对比各种不同存储方式 # 每个表写入5000行记录，观察最后表空间文件大小对比
#
#表0：所有数据存储在一个BLOB列中
CREATE TABLE `t_longcol_0` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol` blob NOT NULL COMMENT 'store all data in a blob column',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ROW_FORMAT=COMPACT;
#相应的数据写入存储过程：mysp_longcol_0_ins()
CREATE PROCEDURE `mysp_longcol_0_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_0(longcol) select repeat('a',65535);
set @i = @i + 1;
end while;
end;
#表1：将64KB字节平均存储在9个列中
CREATE TABLE `t_longcol_1` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` blob NOT NULL COMMENT 'store all data in 9 blob columns',
`longcol2` blob NOT NULL,
`longcol3` blob NOT NULL,
`longcol4` blob NOT NULL,
`longcol5` blob NOT NULL,
`longcol6` blob NOT NULL,
`longcol7` blob NOT NULL,
`longcol8` blob NOT NULL,
`longcol9` blob NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_1_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_1(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9) select
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',5535);
set @i = @i + 1;
end while;
end;
#表2：将64KB数据离散存储在多个BLOB列中
CREATE TABLE `t_longcol_2` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol2` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol3` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol4` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol5` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol6` blob NOT NULL COMMENT 'store 255 bytes data',
`longcol7` blob NOT NULL COMMENT 'store 368 bytes data',
`longcol8` blob NOT NULL COMMENT 'store 496 bytes data',
`longcol9` blob NOT NULL COMMENT 'store 512 bytes data',
`longcol10` blob NOT NULL COMMENT 'store 640 bytes data',
`longcol11` blob NOT NULL COMMENT 'store 768 bytes data',
`longcol12` blob NOT NULL COMMENT 'store 912 bytes data',
`longcol13` blob NOT NULL COMMENT 'store 1024 bytes data',
`longcol14` blob NOT NULL COMMENT 'store 2048 bytes data',
`longcol15` blob NOT NULL COMMENT 'store 3082 bytes data',
`longcol16` blob NOT NULL COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` blob NOT NULL COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_1_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_2(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表3：将64KB数据离散存储在多个CHAR、VARCHAR、BLOB列中
CREATE TABLE `t_longcol_3` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_3_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_3(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表4：将64KB数据离散存储在多个VARCHAR、BLOB列中，对比t_longcol_3中几个列是CHAR的情况
CREATE TABLE `t_longcol_4` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_4_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_4(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表5：将64KB数据离散存储在多个VARCHAR、BLOB列中，和t_longcol_4相比，变化在于前面的几个列长度改成了255，但实际存储长度还是100字节
CREATE TABLE `t_longcol_5` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_5_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_5(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#从下面开始，参考第3条建议进行分表，每个表所有列长度总和
#分表1，行最大长度 100 + 100 + 100 + 100 + 100 + 255 + 368 + 496 + 512 + 640 + 768 + 912 + 3082 = 7533 字节
CREATE TABLE `t_longcol_51` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表2，行最大长度 1024 + 2048 + 4096 = 7168 字节
CREATE TABLE `t_longcol_52` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表3，行最大长度 8192 字节
CREATE TABLE `t_longcol_53` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表4，行最大长度 16284 + 20380 = 36664 字节
CREATE TABLE `t_longcol_54` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表5，行最大长度 5977 + 4 = 5981 字节
CREATE TABLE `t_longcol_55` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_51_ins()
CREATE PROCEDURE `mysp_longcol_51_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_51(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol15) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',3082);
insert into t_longcol_52(longcol13,longcol14,longcol16) select
repeat('a',1024),
repeat('a',2048),
repeat('a',4096);
insert into t_longcol_53(longcol17) select repeat('a',8192);
insert into t_longcol_54(longcol18,longcol19) select
repeat('a',16284),
repeat('a',20380);
insert into t_longcol_55(longcol20) select repeat('a',5977);
set @i = @i + 1;
end while;
end;
上述各个测试表都写入5000行记录后，再来对比下其表空间文件大小，以及重整表空间后的大小，观察碎片率。详细对比见下：
最后一种分表方式中，5个子表的表空间文件大小总和是 40960 + 40960 + 98304 + 286720 + 40960 = 507904 字节。
可以看到，这种方式的总大小和原始表大小差距最小，其他几种存储方式都比这个来的大。
(责任编辑：IT)
首先，介绍下关于InnoDB引擎存储格式的几个要点：
综上，如果在实际业务中，确实需要在InnoDB表中存储BLOB、TEXT、长VARCHAR列时，有下面几点建议： 下面是测试验证过程，有耐心的同学可以慢慢看：
#
# 测试案例：InnoDB中长列存储效率
# 测试场景描述：
# 在InnoDB表中存储64KB的数据，对比各种不同存储方式 # 每个表写入5000行记录，观察最后表空间文件大小对比
#
#表0：所有数据存储在一个BLOB列中
CREATE TABLE `t_longcol_0` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol` blob NOT NULL COMMENT 'store all data in a blob column',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ROW_FORMAT=COMPACT;
#相应的数据写入存储过程：mysp_longcol_0_ins()
CREATE PROCEDURE `mysp_longcol_0_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_0(longcol) select repeat('a',65535);
set @i = @i + 1;
end while;
end;
#表1：将64KB字节平均存储在9个列中
CREATE TABLE `t_longcol_1` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` blob NOT NULL COMMENT 'store all data in 9 blob columns',
`longcol2` blob NOT NULL,
`longcol3` blob NOT NULL,
`longcol4` blob NOT NULL,
`longcol5` blob NOT NULL,
`longcol6` blob NOT NULL,
`longcol7` blob NOT NULL,
`longcol8` blob NOT NULL,
`longcol9` blob NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_1_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_1(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9) select
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',7500),
repeat('a',5535);
set @i = @i + 1;
end while;
end;
#表2：将64KB数据离散存储在多个BLOB列中
CREATE TABLE `t_longcol_2` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol2` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol3` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol4` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol5` blob NOT NULL COMMENT 'store 100 bytes data',
`longcol6` blob NOT NULL COMMENT 'store 255 bytes data',
`longcol7` blob NOT NULL COMMENT 'store 368 bytes data',
`longcol8` blob NOT NULL COMMENT 'store 496 bytes data',
`longcol9` blob NOT NULL COMMENT 'store 512 bytes data',
`longcol10` blob NOT NULL COMMENT 'store 640 bytes data',
`longcol11` blob NOT NULL COMMENT 'store 768 bytes data',
`longcol12` blob NOT NULL COMMENT 'store 912 bytes data',
`longcol13` blob NOT NULL COMMENT 'store 1024 bytes data',
`longcol14` blob NOT NULL COMMENT 'store 2048 bytes data',
`longcol15` blob NOT NULL COMMENT 'store 3082 bytes data',
`longcol16` blob NOT NULL COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` blob NOT NULL COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_1_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_2(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表3：将64KB数据离散存储在多个CHAR、VARCHAR、BLOB列中
CREATE TABLE `t_longcol_3` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` char(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_3_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_3(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表4：将64KB数据离散存储在多个VARCHAR、BLOB列中，对比t_longcol_3中几个列是CHAR的情况
CREATE TABLE `t_longcol_4` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(100) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_4_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_4(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#表5：将64KB数据离散存储在多个VARCHAR、BLOB列中，和t_longcol_4相比，变化在于前面的几个列长度改成了255，但实际存储长度还是100字节
CREATE TABLE `t_longcol_5` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_5_ins()
CREATE PROCEDURE `mysp_longcol_1_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_5(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol13,longcol14,longcol15,longcol16,longcol17,longcol18,longcol19,longcol20) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',1024),
repeat('a',2048),
repeat('a',3082),
repeat('a',4096),
repeat('a',8192),
repeat('a',16284),
repeat('a',20380),
repeat('a',5977);
set @i = @i + 1;
end while;
end;
#从下面开始，参考第3条建议进行分表，每个表所有列长度总和
#分表1，行最大长度 100 + 100 + 100 + 100 + 100 + 255 + 368 + 496 + 512 + 640 + 768 + 912 + 3082 = 7533 字节
CREATE TABLE `t_longcol_51` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol1` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol2` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol3` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol4` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol5` varchar(255) NOT NULL DEFAULT '' COMMENT 'store 100 bytes data',
`longcol6` varchar(256) NOT NULL DEFAULT '' COMMENT 'store 255 bytes data',
`longcol7` varchar(368) NOT NULL DEFAULT '' COMMENT 'store 368 bytes data',
`longcol8` varchar(496) NOT NULL DEFAULT '' COMMENT 'store 496 bytes data',
`longcol9` varchar(512) NOT NULL DEFAULT '' COMMENT 'store 512 bytes data',
`longcol10` varchar(640) NOT NULL DEFAULT '' COMMENT 'store 640 bytes data',
`longcol11` varchar(768) NOT NULL DEFAULT '' COMMENT 'store 768 bytes data',
`longcol12` varchar(912) NOT NULL DEFAULT '' COMMENT 'store 912 bytes data',
`longcol15` varchar(3082) NOT NULL DEFAULT '' COMMENT 'store 3082 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表2，行最大长度 1024 + 2048 + 4096 = 7168 字节
CREATE TABLE `t_longcol_52` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol13` varchar(1024) NOT NULL DEFAULT '' COMMENT 'store 1024 bytes data',
`longcol14` varchar(2048) NOT NULL DEFAULT '' COMMENT 'store 2048 bytes data',
`longcol16` varchar(4096) NOT NULL DEFAULT '' COMMENT 'store 4096 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表3，行最大长度 8192 字节
CREATE TABLE `t_longcol_53` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol17` blob NOT NULL COMMENT 'store 8192 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表4，行最大长度 16284 + 20380 = 36664 字节
CREATE TABLE `t_longcol_54` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol18` blob NOT NULL COMMENT 'store 16284 bytes data',
`longcol19` blob NOT NULL COMMENT 'store 20380 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#分表5，行最大长度 5977 + 4 = 5981 字节
CREATE TABLE `t_longcol_55` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`longcol20` varchar(5977) NOT NULL DEFAULT '' COMMENT 'store 5977 bytes data',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
#相应的数据写入存储过程：mysp_longcol_51_ins()
CREATE PROCEDURE `mysp_longcol_51_ins`( in cnt int )
begin
set @i = 1;
while @i < cnt do
insert into t_longcol_51(longcol1,longcol2,longcol3,longcol4,longcol5,longcol6,longcol7,longcol8,longcol9,longcol10,
longcol11,longcol12,longcol15) select
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',100),
repeat('a',256),
repeat('a',368),
repeat('a',496),
repeat('a',512),
repeat('a',640),
repeat('a',768),
repeat('a',912),
repeat('a',3082);
insert into t_longcol_52(longcol13,longcol14,longcol16) select
repeat('a',1024),
repeat('a',2048),
repeat('a',4096);
insert into t_longcol_53(longcol17) select repeat('a',8192);
insert into t_longcol_54(longcol18,longcol19) select
repeat('a',16284),
repeat('a',20380);
insert into t_longcol_55(longcol20) select repeat('a',5977);
set @i = @i + 1;
end while;
end;
上述各个测试表都写入5000行记录后，再来对比下其表空间文件大小，以及重整表空间后的大小，观察碎片率。详细对比见下：
最后一种分表方式中，5个子表的表空间文件大小总和是 40960 + 40960 + 98304 + 286720 + 40960 = 507904 字节。 |