LOAD DATA INFILE – performance case study
转:
http://venublog.com/2007/11/07/load-data-infile-performance/
I often noticed that people complain about the LOAD DATA performance when loading the table with large number of rows of data. Even today I saw a case where the LOAD DATA on a simple 3 column table with about 5 million rows taking ~15 minutes of time. This is because the server did not had any tuning in regards to bulk insertion.
Consider the following simple MyISAM table on Redhat Linux 32-bit.
Shell
|
1
2
3
4
5
6
7
8
|
CREATE TABLE load1 (
`col1` varchar(100) NOT NULL default '',
`col2` int(11) default NULL,
`col3` char(1) default NULL,
PRIMARY KEY (`col1`)
) TYPE=MyISAM;
|
The table has a string key column. Here is the data file(download here) that I used it for testing:
Shell
|
1
2
3
4
5
6
7
|
[vanugant@escapereply:t55 tmp]$ wc loaddata.csv
5164946 5164946 227257389 loaddata.csv
[vanugant@escapereply:t55 tmp]$ ls -alh loaddata.csv
-rw-r--r-- 1 vanugant users 217M Nov 6 14:42 loaddata.csv
[vanugant@escapereply:t55 tmp]$
|
Here is the default mysql system variables related to LOAD DATA:
Shell
|
1
2
3
4
5
6
7
8
9
10
|
mysql> show variables;
+-------------------------+---------+
| Variable_name | Value |
+-------------------------+---------+
| bulk_insert_buffer_size | 8388608 |
| myisam_sort_buffer_size | 16777216 |
| key_buffer_size | 33554432 |
+-------------------------+----------+
|
and here is the actual LOAD DATA query to load all ~5m rows (~256M of data) to the table and its timing.
Shell
|
1
2
3
4
5
|
mysql> LOAD DATA INFILE '/home/vanugant/tmp/loaddata.csv' IGNORE INTO TABLE load1 FIELDS TERMINATED BY ',';
Query OK, 4675823 rows affected (14 min 56.84 sec)
Records: 5164946 Deleted: 0 Skipped: 489123 Warnings: 0
|
Now, lets experiment by disabling the keys in the table before running the LOAD DATA:
Shell
|
1
2
3
4
5
6
7
8
9
10
11
|
mysql> SET SESSION BULK_INSERT_BUFFER_SIZE=314572800;
Query OK, 0 rows affected (0.00 sec)
mysql> alter table load1 disable keys;
Query OK, 0 rows affected (0.00 sec)
mysql> LOAD DATA INFILE '/home/vanugant/tmp/loaddata.csv' IGNORE INTO TABLE load1 FIELDS TERMINATED BY ',';
Query OK, 4675823 rows affected (13 min 47.50 sec)
Records: 5164946 Deleted: 0 Skipped: 489123 Warnings: 0
|
No use, just 1% increase or same…., now lets set the real MyISAM values… and try again…
Shell
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
mysql> SET SESSION BULK_INSERT_BUFFER_SIZE=256217728;
Query OK, 0 rows affected (0.00 sec)
mysql> set session MYISAM_SORT_BUFFER_SIZE=256217728;
Query OK, 0 rows affected (0.00 sec)
mysql> set global KEY_BUFFER_SIZE=256217728;
Query OK, 0 rows affected (0.05 sec)
mysql> alter table load1 disable keys;
Query OK, 0 rows affected (0.00 sec)
mysql> LOAD DATA INFILE '/home/vanugant/tmp/loaddata.csv' IGNORE INTO TABLE load1 FIELDS TERMINATED BY ',';
Query OK, 4675823 rows affected (1 min 55.05 sec)
Records: 5164946 Deleted: 0 Skipped: 489123 Warnings: 0
mysql> alter table load1 enable keys;
Query OK, 0 rows affected (0.00 sec)
|
Wow…thats almost 90% increase in the performance. So, disabling the keys in MyISAM is not just the key, but tuning the buffer size does play role based on the input data.
For the same case with Innodb, here is the status by adjusting the Innodb_buffer_pool_size=1G andInnodb_log_file_size=256M along with innodb_flush_logs_at_trx_commit=1.
Shell
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
|
mysql> show variables like '%innodb%size';
+---------------------------------+------------+
| Variable_name | Value |
+---------------------------------+------------+
| innodb_additional_mem_pool_size | 26214400 |
| innodb_buffer_pool_size | 1073741824 |
| innodb_log_buffer_size | 8388608 |
| innodb_log_file_size | 268435456 |
+---------------------------------+------------+
mysql> LOAD DATA INFILE '/home/vanugant/tmp/loaddata.csv' IGNORE INTO TABLE load1 FIELDS TERMINATED BY ',';
Query OK, 4675823 rows affected (2 min 37.53 sec)
Records: 5164946 Deleted: 0 Skipped: 489123 Warnings: 0
|
With innodb_flush_logs_at_trx_commit=2, innodb_flush_method=O_DIRECT and innodb_doublewrite=0; it will be another 40% difference (use all these variables with caution, unless you know what you are doing)
Shell
|
1
2
3
4
5
|
mysql> LOAD DATA INFILE '/home/vanugant/tmp/loaddata.csv' IGNORE INTO TABLE load1 FIELDS TERMINATED BY ',';
Query OK, 4675823 rows affected (1 min 53.69 sec)
Records: 5164946 Deleted: 0 Skipped: 489123 Warnings: 0
|
LOAD DATA INFILE – performance case study的更多相关文章
- LOAD DATA INFILE Syntax--官方
LOAD DATA [LOW_PRIORITY | CONCURRENT] [LOCAL] INFILE 'file_name' [REPLACE | IGNORE] INTO TABLE tbl_n ...
- Data Visualization – Banking Case Study Example (Part 1-6)
python信用评分卡(附代码,博主录制) https://study.163.com/course/introduction.htm?courseId=1005214003&utm_camp ...
- Mysql load data infile 导入数据出现:Data truncated for column
[1]Mysql load data infile 导入数据出现:Data truncated for column .... 可能原因分析: (1)数据库表对应字段类型长度不够或修改为其他数据类型( ...
- Mysql load data infile 命令导入含中文csv源数据文件 【错误代码 1300】
[1]Load data infile 命令导入含中文csv源数据文件 报错:Invalid utf8 character string: '??֧' (1)问题现象 csv格式文件源数据: 导入SQ ...
- Mysql load data infile 命令格式
[1]Linux系统环境下 LOAD DATA INFILE /usr/LOCAL/lib/ubcsrvd/datacsv/201909_source.csv INTO TABLE np_cdr_20 ...
- Mysql 命令 load data infile 权限问题
[1]Mysql命令load data infile 执行权限问题 工作中,经常会遇到往线上环境mysql数据库批量导入源数据的场景. 针对这个场景问题,mysql有一个很高效的命令:load dat ...
- mysql load data infile的使用 和 SELECT into outfile备份数据库数据
LOAD DATA [LOW_PRIORITY | CONCURRENT] [LOCAL] INFILE 'file_name.txt' [REPLACE | IGNORE] INTO TABLE t ...
- SQL基本语句(3) LOAD DATA INFILE
使用LOAD语句批量录入数据 语法: LOAD DATA [LOW_PRIORITY | CONCURRENT] [LOCAL] INFILE 'file_name' [REPLACE | IGNOR ...
- mysql导入数据load data infile用法
mysql导入数据load data infile用法 基本语法: load data [low_priority] [local] infile 'file_name txt' [replace | ...
随机推荐
- Linux Shell查看磁盘分区,内存使用,CPU使用率
Linux Shell查看磁盘分区,内存使用,CPU使用率 #!/bin/bash #disk_used_rate Location=/dev/xvdb Disk_Used_Rate=$(df -h ...
- [iOS]集成环信SDK然后运行时候crash了-[NSBundle initWithURL:]: nil URL argument'
Crash的reason是-[NSBundle initWithURL:]: nil URL argument' 1.首先我是用cocoapods导入的环信的SDK.然后怎么运行怎么crash. 2. ...
- Spring MVC配置DispatcherServlet的url-pattern
在配置Spring MVC的核心过滤器DispatcherServlet的url-pattern时是有要求的. <servlet> <servlet-name>...</ ...
- Oracle ->> 随机函数
SQL SERVER下生成随机数据干得多,可是到了Oracle下我就傻了.没用过Oracle,不知道该怎么办?SQL SERVER下依靠TABLESAMPLE或者CHECKSUM(NEWID())来做 ...
- 浅谈Linux下的五种I/O模型
一.关于I/O模型的引出 我们都知道,为了OS的安全性等的考虑,进程是无法直接操作I/O设备的,其必须通过系统调用请求内核来协助完成I/O动作,而内核会为每个I/O设备维护一个buffer.如下图所 ...
- CMD下查看某个端口被谁占用了
cmd运行 netstat -aon|findstr "8080" 运行结果如下 TCP 0.0.0.0:8080 0.0.0.0:0 LISTENING 6416 TCP [:: ...
- 错误 -force-32bit 与 ANDROID_EMULATOR_FORCE_32BIT=true
1,配置环境变量, 加上ANDROID_EMULATOR_FORCE_32BIT=true 2,在AS中启动模拟器用下面方法 在你要运行的个工程右击->Run as -> Run conf ...
- Evaluate Reverse Polish Notation(堆栈)
Evaluate the value of an arithmetic expression in Reverse Polish Notation. Valid operators are +, -, ...
- VS2015新功能
今天有幸参加了微软的 Visual Studio Dev Day,趁还没有忘记今天的学习内容. 先把这些内容记录下来,如果有其他人也参加此次交流活动,请补充完善. VS2015新功能 1,Roslyn ...
- linux添加环境变量
修改 /当前用户/.bash_profile文件,修改export PATH,添加自己的可执行程序的目录即可 例如: /root/.bash_profile export PATH=$PATH:/u ...