背景:

      TokuDB引擎是有Tokutek开发的一个数据库存储引擎,在设计之初便引入了独特的索引算法,在其官网测试的文章中看到TokuDB性能比InnoDB高出很多。

      MySQL是一个插件式的数据库,在MySQL5.5版本之前MyISAM是MySQL的默认存储引擎,在之后的版本中默认的存储引擎变成了InnoDB。其特点是它支持事务,具有完善的崩溃恢复机制,具体的特点这里不说明,可以自行的寻找资料。这里介绍的TokuDB和InnoDB有很多相似之处:一个高性能,支持事务、MVCC、聚簇索引等。最大的不同在于TokuDB采用了一种叫做Fractal Tree的索引结构,使其在随机写数据的处理上有很大提升。一般来说数据库的索引结构都采用B+Tree或则类似的数据结构,InnoDB也是如此。InnoDB是以主键组织的B+Tree结构,数据按照主键顺序排列。对于顺序的自增主键有很好的性能,但是不适合随机写入,大量的随机I/O会使数据页分裂产生碎片,索引维护开销很多大。而TokuDB的Fractal Tree的索引结构很好的解决了这个问题。

      TokuDB解决随机写入的问题得益于其索引结构,Fractal Tree 和 B-Tree的差别主要在于索引树的内部节点上,B-Tree索引的内部结构只有指向父节点和子节点的指针,而Fractal Tree的内部节点不仅有指向父节点和子节点的指针,还有一块Buffer区。当数据写入时会先落到这个Buffer区上,该区是一个FIFO结构,写是一个顺序的过程,和其他缓冲区一样,满了就一次性刷写数据。所以TokuDB上插入数据基本上变成了一个顺序添加的过程。

      TokuDB另一个特点是压缩性能和低CPU消耗,TokuDB存储引擎默认的块大小是4M,这使其有更好的压缩效率。默认支持压缩功能,不需要配置其他的东西。压缩选项有:TokuDB_Quicklz、TokuDB_Lzma、TokuDB_Zlib,同时也支持非压缩选项。TokuDB_Zlib支持的默认压缩格式。一般压缩都需要消耗更多的CPU。但TukuDB消耗的CPU资源较少。

TokuDBhttps://www.percona.com/doc/percona-server/5.7/index.html

安装

      为了方便安装,直接在percona版本的MySQL上安装TokuDB存储引擎。Ubuntu下apt安装percona MySQL:http://www.percona.com/doc/percona-server/5.6/installation/apt_repo.html

vi /etc/apt/sources.list.d/percona.list

deb http://repo.percona.com/apt trusty main

其他:添加其他软件的更新,和本次安装无关

vi /etc/apt/sources.list

deb http://mirrors.aliyun.com/ubuntu trusty main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu trusty-updates main restricted universe multiverse
deb http://security.ubuntu.com/ubuntu trusty-security main restricted universe multiverse

源添加完毕之后更新:

apt-get update

最后apt安装percona MySQL:

apt-get install percona-server-server-5.6 percona-server-client-5.6

以上percona MySQL安装完毕,接着就开始安装TokuDB。

参考官网的安装说明:https://www.percona.com/doc/percona-server/5.6/tokudb/tokudb_installation.html

安装TokuDB的前提条件是需要支持jemalloc,需要关闭 Transparent huge pages ,执行:

echo never > /sys/kernel/mm/transparent_hugepage/enabled
echo never > /sys/kernel/mm/transparent_hugepage/defrag

效果:

root@ptest:~# cat /sys/kernel/mm/transparent_hugepage/enabled
always madvise [never]
root@ptest:~# cat /sys/kernel/mm/transparent_hugepage/defrag
always madvise [never]

开始apt安装TokuDB:

root@ptest:~# apt-get install percona-server-tokudb-5.6

安装好检查:

root@ptest:~# ps_tokudb_admin --enable -uroot -p
Enter password:
Continuing without password...


Checking if Percona server is running with jemalloc enabled...
grep: /proc/5440/environ: Permission denied
>> Percona server is running with jemalloc enabled.

Checking transparent huge pages status on the system...
>> Transparent huge pages are currently disabled on the system.

Checking if thp-setting=never option is already set in config file...
>> Option thp-setting=never is not set in the config file.
>> (needed only if THP is not disabled permanently on the system)

Checking TokuDB plugin status...
>> TokuDB plugin is installed.

Adding thp-setting=never option into /etc/mysql/my.cnf
>> Successfuly added thp-setting=never option into /etc/mysql/my.cnf   #把参数thp-setting=never加载到了my.cnf的[mysqld_safe]的选项组下

通过show engines、show plugins 查看存储引擎,是否安装成功,要是没有成功则检查上面是否有错误。要是通过show engines没有TokuDB,但通过show plugins有则需要手动install:

INSTALL PLUGIN tokudb SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_file_map SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_fractal_tree_info SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_fractal_tree_block_map SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_trx SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_locks SONAME 'ha_tokudb.so';
INSTALL PLUGIN tokudb_lock_waits SONAME 'ha_tokudb.so';

查看版本:

>SELECT @@tokudb_version;
+------------------+
| @@tokudb_version |
+------------------+
| tokudb-7.5.6     |
+------------------+

以上就完成了TokuDB的安装。详细信息请见官网

测试:针对插入、压缩、存储引擎转换、事务、复制等日常用到的功能进行测试。

1)插入和压缩性能:压缩参数:tokudb_row_format

zjy@127.0.0.1 : test 10:58:58>show create table login_log_Toku\G;
*************************** 1. row ***************************
       Table: login_log_Toku
Create Table: CREATE TABLE `login_log_Toku` (
  `userID` varchar(255) NOT NULL,
  `tryIP` varchar(255) DEFAULT NULL,
  `tryTime` datetime DEFAULT NULL,
  `tryResult` tinyint(1) DEFAULT '0',
  `refferUrl` varchar(255) DEFAULT NULL,
  `sID` int(10) unsigned NOT NULL AUTO_INCREMENT,
  `authType` tinyint(1) DEFAULT '1',
  `type` tinyint(1) DEFAULT '1',
  `deviceType` varchar(64) DEFAULT '',
  PRIMARY KEY (`sID`),
  KEY `idx_tryip` (`tryIP`),
  KEY `idx_userid` (`userID`),
  KEY `idx_tryTime` (`tryTime`)
) ENGINE=TokuDB DEFAULT CHARSET=utf8
1 row in set (0.01 sec)

ERROR:
No query specified

zjy@127.0.0.1 : test 10:59:45>insert into login_log_Toku select * from login_log;
Query OK, 49534274 rows affected (18 min 35.63 sec)
Records: 49534274  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 11:18:40>create table login_log_innodb like login_log;
Query OK, 0 rows affected (0.11 sec)

zjy@127.0.0.1 : test 11:27:58>insert into login_log_innodb select * from login_log;
Query OK, 49534274 rows affected (2 hours 1 min 34.06 sec)
Records: 49534274  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 01:29:38>create table login_log_myisam like login_log;
Query OK, 0 rows affected (0.15 sec)

zjy@127.0.0.1 : test 08:56:30>alter table login_log_myisam engine = myisam;
Query OK, 0 rows affected (0.10 sec)
Records: 0  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 08:56:51>insert into login_log_myisam select * from login_log;
Query OK, 49534274 rows affected (14 min 25.44 sec)
Records: 49534274  Duplicates: 0  Warnings: 0

上面看到写入TokuDB表的时间远远小于InnoDB,MyISAM(堆表)的写入也比InnoDB好很多。

-rw-rw---- 1 mysql mysql 8.7K  5月  7 23:27 login_log_innodb.frm
-rw-rw---- 1 mysql mysql 9.7G  5月  8 03:12 login_log_innodb.ibd
-rw-rw---- 1 mysql mysql 8.7K  5月  8 08:56 login_log_myisam.frm
-rw-rw---- 1 mysql mysql 3.8G  5月  8 09:00 login_log_myisam.MYD
-rw-rw---- 1 mysql mysql 1.7G  5月  8 09:11 login_log_myisam.MYI
-rw-rw---- 1 mysql mysql 8.7K  5月  8 16:43 login_log_Toku.frm

查看文件,看到InnoDB所占的空间比MyISAM大,9.7>5.5。TokuDB在数据库目录下面只保存.frm表结构定义文件,其数据保存在数据库的目录的上一层(可以通过tokudb_data_dir,tokudb_log_dir指定位置):

ls -lh *toku*
-rw------- 1 mysql mysql  12K  5月 12 17:00 log000000000005.tokulog27
-rw-rw---- 1 mysql mysql  64K  5月  8 16:43 _test_sql_c80_2a_status_34_1_1b.tokudb
-rw-rw---- 1 mysql mysql 265M  5月  8 16:43 _test_login_log_Toku_key_idx_tryip_3c_1_1b_B_1.tokudb
-rw-rw---- 1 mysql mysql 229M  5月  8 16:43 _test_login_log_Toku_key_idx_tryTime_3c_1_1b_B_3.tokudb
-rw-rw---- 1 mysql mysql 282M  5月  8 16:43 _test_login_log_Toku_key_idx_userid_3c_1_1b_B_2.tokudb
-rw-rw---- 1 mysql mysql 1.4G  5月  8 18:25 _test_login_log_Toku_main_3c_1_1b_B_0.tokudb
-rw-rw---- 1 mysql mysql  17K  5月  7 17:10 _test_ttt_main_5_1_1b_B_0.tokudb
-rw-rw---- 1 mysql mysql  16K  5月  7 17:10 _test_ttt_status_3_1_1b.tokudb
-rw-rw---- 1 mysql mysql  32K  5月  8 16:43 tokudb.directory
-rw-rw---- 1 mysql mysql  16K  5月  7 17:09 tokudb.environment
-rw------- 1 mysql mysql    0  5月  7 17:09 __tokudb_lock_dont_delete_me_data
-rw------- 1 mysql mysql    0  5月  7 17:09 __tokudb_lock_dont_delete_me_environment
-rw------- 1 mysql mysql    0  5月  7 17:09 __tokudb_lock_dont_delete_me_logs
-rw------- 1 mysql mysql    0  5月  7 17:09 __tokudb_lock_dont_delete_me_recovery
-rw------- 1 mysql mysql    0  5月  7 17:09 __tokudb_lock_dont_delete_me_temp
-rw-rw---- 1 mysql mysql  32K  5月  8 16:43 tokudb.rollback

从上面看出,TokuDB的索引文件和数据文件是分开的,一个索引一个文件。总共大小为2.2G,最小。其他为TokuDB的日志文件和一些元数据。

2)存储引擎转换和事务支持:

1:存储引擎转换

zjy@127.0.0.1 : test 06:08:48>show create table log\G;
*************************** 1. row ***************************
       Table: log
Create Table: CREATE TABLE `log` (
  `userID` varchar(255) NOT NULL,
  `tryIP` varchar(255) DEFAULT NULL,
  `tryTime` datetime DEFAULT NULL,
  `tryResult` tinyint(1) DEFAULT '0',
  `refferUrl` varchar(255) DEFAULT NULL,
  `sID` int(10) unsigned NOT NULL DEFAULT '0',
  `authType` tinyint(1) DEFAULT '1',
  `type` tinyint(1) DEFAULT '1',
  `deviceType` varchar(64) DEFAULT '',
  KEY `idx_userid` (`userID`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8
1 row in set (0.01 sec)

ERROR: 
No query specified

zjy@127.0.0.1 : test 06:08:50>alter table log engine = tokudb;
Query OK, 100 rows affected (0.02 sec)
Records: 100  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 06:08:56>select * from log;
有数据,数据未丢失。

zjy@127.0.0.1 : test 06:09:34>show create table log\G;
*************************** 1. row ***************************
       Table: log
Create Table: CREATE TABLE `log` (
  `userID` varchar(255) NOT NULL,
  `tryIP` varchar(255) DEFAULT NULL,
  `tryTime` datetime DEFAULT NULL,
  `tryResult` tinyint(1) DEFAULT '0',
  `refferUrl` varchar(255) DEFAULT NULL,
  `sID` int(10) unsigned NOT NULL DEFAULT '0',
  `authType` tinyint(1) DEFAULT '1',
  `type` tinyint(1) DEFAULT '1',
  `deviceType` varchar(64) DEFAULT '',
  KEY `idx_userid` (`userID`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8
1 row in set (0.12 sec)

ERROR: 
No query specified

zjy@127.0.0.1 : test 06:09:37>alter table log engine = tokudb;
Query OK, 100 rows affected (0.02 sec)
Records: 100  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 06:15:24>select * from log;
有数据,数据未丢失。

反推把TokuDB转换成InnoDB和MyISAM也没有问题。

2:事务支持

zjy@127.0.0.1 : test 03:20:19>create table tmp_innodb(id int,name varchar(10))engine = innodb default charset utf8;
Query OK, 0 rows affected (0.44 sec)

zjy@127.0.0.1 : test 03:20:57>create table tmp_myisam(id int,name varchar(10))engine = myisam default charset utf8;
Query OK, 0 rows affected (0.01 sec)

zjy@127.0.0.1 : test 03:21:07>create table tmp_tokudb(id int,name varchar(10))engine = tokudb default charset utf8;
Query OK, 0 rows affected (0.05 sec)

zjy@127.0.0.1 : test 03:21:17>start transaction;
Query OK, 0 rows affected (0.01 sec)

zjy@127.0.0.1 : test 03:21:47>insert into tmp_innodb values(1,'a'),(2,'b'),(3,'c');
Query OK, 3 rows affected (0.11 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:22:10>insert into tmp_myisam values(1,'a'),(2,'b'),(3,'c');
Query OK, 3 rows affected (0.01 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:22:17>insert into tmp_tokudb values(1,'a'),(2,'b'),(3,'c');
Query OK, 3 rows affected (0.01 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:22:22>commit;
Query OK, 0 rows affected (0.02 sec)

zjy@127.0.0.1 : test 03:22:26>select * from tmp_innodb;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.16 sec)

zjy@127.0.0.1 : test 03:22:52>select * from tmp_myisam;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.01 sec)

zjy@127.0.0.1 : test 03:22:56>select * from tmp_tokudb;    #和innodb一样,数据提交写入
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.01 sec)

zjy@127.0.0.1 : test 03:23:04>start transaction;                                                                                       Query OK, 0 rows affected (0.16 sec)

zjy@127.0.0.1 : test 03:23:17>insert into tmp_innodb values(11,'aa'),(22,'bb'),(33,'cc');
Query OK, 3 rows affected (0.01 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:23:39>insert into tmp_myisam values(11,'aa'),(22,'bb'),(33,'cc');
Query OK, 3 rows affected (0.12 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:23:51>insert into tmp_tokudb values(11,'aa'),(22,'bb'),(33,'cc');
Query OK, 3 rows affected (0.00 sec)
Records: 3  Duplicates: 0  Warnings: 0

zjy@127.0.0.1 : test 03:24:30>rollback;
Query OK, 0 rows affected, 1 warning (0.01 sec)   #myisam不支持事务引发的warning

zjy@127.0.0.1 : test 03:24:40>select * from tmp_innodb;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.29 sec)

zjy@127.0.0.1 : test 03:24:49>select * from tmp_myisam;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
|   11 | aa   |
|   22 | bb   |
|   33 | cc   |
+------+------+
6 rows in set (0.18 sec)

zjy@127.0.0.1 : test 03:24:55>select * from tmp_tokudb;  #和innodb一样,数据被回滚。
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|    3 | c    |
+------+------+
3 rows in set (0.00 sec)

3)复制相关

       MySQL的复制是通过Master利用binlog IO线程把修改数据库的SQL推送到Slave上,Slave通过本地的binlog IO线程把SQL写入到中继日志里(relay log)里,再通过SQL线程回放这些SQL,应用到数据库。复制的核心在于binlog,而binlog和存储引擎是没有关系,所以从理论上说主从的存储引擎不一样也不影响复制。

MASTER:主上建表写入
rep@192.168.200.25 : test 08:55:58>create table tmp_engine(id int,name varchar(10))default charset utf8;
Query OK, 0 rows affected (0.24 sec)

rep@192.168.200.25 : test 08:56:57>insert into tmp_engine values(1,'a'),(2,'b');
Query OK, 2 rows affected (0.15 sec)
Records: 2  Duplicates: 0  Warnings: 0

rep@192.168.200.25 : test 08:59:18>insert into tmp_engine values(11,'aa'),(22,'bb');
Query OK, 2 rows affected (0.37 sec)
Records: 2  Duplicates: 0  Warnings: 0

rep@192.168.200.25 : test 08:59:25>select * from tmp_engine;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|   11 | aa   |
|   22 | bb   |
+------+------+
4 rows in set (0.24 sec)

SLAVE:从上同步数据,并修改存储引擎
zjy@127.0.0.1 : test 09:00:01>select * from tmp_engine;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|   11 | aa   |
|   22 | bb   |
+------+------+
4 rows in set (0.19 sec)

zjy@127.0.0.1 : test 09:00:04>alter table tmp_engine engine = tokudb;
Query OK, 4 rows affected (0.02 sec)
Records: 4  Duplicates: 0  Warnings: 0

MASTER:主上继续写入
rep@192.168.200.25 : test 09:00:24>insert into tmp_engine values(1111,'aaaa'),(2222,'bbbb');
Query OK, 2 rows affected (0.02 sec)
Records: 2  Duplicates: 0  Warnings: 0

rep@192.168.200.25 : test 09:00:29>select * from tmp_engine;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|   11 | aa   |
|   22 | bb   |
|  111 | aaa  |
|  222 | bbb  |
| 1111 | aaaa |
| 2222 | bbbb |
+------+------+
8 rows in set (0.05 sec)

SLAVE:从上修改完存储引擎之后,数据继续正常同步。
zjy@127.0.0.1 : test 09:00:13>select * from tmp_engine;
+------+------+
| id   | name |
+------+------+
|    1 | a    |
|    2 | b    |
|   11 | aa   |
|   22 | bb   |
|  111 | aaa  |
|  222 | bbb  |
| 1111 | aaaa |
| 2222 | bbbb |
+------+------+
8 rows in set (0.23 sec)

从上面看出,主从复制和存储引擎无关,从上修改成和主上不同的存储引擎,正常同步。

4)备份相关

TokuDB 有自己的热备份方案:tokudb-hot-backup-part-1tokudb-hot-backup-part-2,和其他存储引擎一样也可以使用mysqldump进行备份。

-- MySQL dump 10.13  Distrib 5.6.19, for osx10.7 (x86_64)
--
-- Host: 192.168.200.240    Database: test
-- ------------------------------------------------------
-- Server version       5.6.23-72.1

/*!40101 SET @OLD_CHARACTER_SET_CLIENT=@@CHARACTER_SET_CLIENT */;
/*!40101 SET @OLD_CHARACTER_SET_RESULTS=@@CHARACTER_SET_RESULTS */;
/*!40101 SET @OLD_COLLATION_CONNECTION=@@COLLATION_CONNECTION */;
/*!40101 SET NAMES utf8 */;
/*!40103 SET @OLD_TIME_ZONE=@@TIME_ZONE */;
/*!40103 SET TIME_ZONE='+00:00' */;
/*!40014 SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0 */;
/*!40014 SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0 */;
/*!40101 SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='NO_AUTO_VALUE_ON_ZERO' */;
/*!40111 SET @OLD_SQL_NOTES=@@SQL_NOTES, SQL_NOTES=0 */;

--
-- Table structure for table `tmp_engine`
--

DROP TABLE IF EXISTS `tmp_engine`;
/*!40101 SET @saved_cs_client     = @@character_set_client */;
/*!40101 SET character_set_client = utf8 */;
CREATE TABLE `tmp_engine` (
  `id` int(11) DEFAULT NULL,
  `name` varchar(10) DEFAULT NULL
) ENGINE=TokuDB DEFAULT CHARSET=utf8;
/*!40101 SET character_set_client = @saved_cs_client */;

--
-- Dumping data for table `tmp_engine`
View Code

相关文章: