Performance optimization record of the company's product "yunzhujia"

Recently, the performance of the company's products has been preliminarily verified and some preliminary optimization has been carried out , The following is a record of this optimization process , It is convenient for you to consult in the future ：

■ Test environment ：

windows server 2008 

8G Memory

JDK1.7

apache2.4.x+tomcat8.0.x

mysql5.6.24

■ Test scenarios ：

sys_info surface 6000 Data

sys_dynamic_share surface 10000 Data

sys_article surface 10000 Data

100 Users access the consultation list page concurrently ;

■ Optimization steps :

1、 Find out the slow query : Through analysis sql sentence , adjustment sql How to write it , At the same time, add the necessary indexes .

Harvest ：（1）mysql adopt explain The table corresponding to the first data returned by the analysis query statement is

Driver table automatically recognized by the database engine , Try to consider the optimization of driving tables when optimizing queries .

（2） Function will invalidate the index .

2、 Verification completed sql After the query performance problem of , Start concurrent validation .

Single user operation , The information list page and information details page respond quickly .

100 After user concurrency , Information list page http The request response time has reached 50 About seconds .

So consider optimizing the configuration environment

（1） According to the server memory , adjustment jvm Related configuration

rem Guess CATALINA_HOME if not defined

set JAVA_OPTS=-Xms2048m -Xmx2048m -XX:PermSize=1024m -XX:MaxNewSize=1024m -XX:MaxPermSize=1024m

After practice , In the process of multi-user concurrency JVM It consumes a lot of heap memory , Therefore, we can properly consider improving the configuration of heap memory .

（2） To configure apache Related parameters

stay apache Of httpd.conf Find

Include conf/extra/httpd-mpm.conf、Include "conf/extra/httpd-default.conf", Remove the annotation symbol after finding ”#“.

→ To conf/extra Found in folder httpd-mpm.conf, Find the following content and modify the value

<IfModule mpm_winnt_module>

    ThreadsPerChild        500

    MaxConnectionsPerChild   5000

</IfModule>

→ To conf/extra Found in folder httpd-default.conf, Also on Timeout、KeepAlive、MaxKeepAliveRequests、KeepAliveTimeout Modify the isoparametric value

Reference materials ： http://www.open-open.com/lib/view/open1388890058296.html

《apache Performance optimization configuration 》：http://blog.csdn.net/hytfly/article/details/17232143

（3） open apache Of GZIP Compression

3、 After modifying the configuration above , Re execute the concurrency test , Find out http The decline of request response time is not obvious .

open mysql Database monitoring , Analysis database , It is found that the queuing phenomenon of data processes is serious , Speculation leads to http The reason for the long response time

It is due to the performance of the database . After investigation and analysis , stay mysql Of my.ini The following configuration is added to the file （ Orange font is the core configuration ）：

key_buffer = 16M

max_allowed_packet = 128M

sort_buffer_size = 512K

net_buffer_length = 8K

read_buffer_size = 256K

read_rnd_buffer_size = 512K

myisam_sort_buffer_size = 8M

log_error = "mysql_error.log"

innodb_buffer_pool_size =3G

max_connections=500

query_cache_type=1

query_cache_size=128M

interactive_timeout=60

wait_timeout=60

innodb_thread_concurrency=64( According to the physical memory configuration ,1G->8,2G->16,3G->32,>4G->64)

notes ： The above configuration needs to be verified according to the actual environment of the server , You can't just copy .

Reference material ：

http://www.cnblogs.com/jiunadianshi/articles/2475475.html

http://www.oschina.net/question/4873_26324（ myql Optimize , start-up mysql Caching mechanisms , Achieve hit rate 100%）

4、 modify jdbc Configuration item ：

Reference material ：《dbcp Basic configuration related configuration description :initialSize： The number of initial connections created when the connection pool starts 》    

http://www.educity.cn/wenda/388309.html

《 Connect... In the pool maxIdle,MaxActive,maxWait Parameters 》

http://blog.csdn.net/xiaxiaorui2003/article/details/7065699

5、memcached Distributed cache

memcached The cache is currently configured in yunzhujia , But there is no real use , Mainly because memcached The cache cannot be updated in time .

On the other hand, it also shows that caching is mainly based on data that rarely changes, such as basic data . Of course, there are many solutions , You can do some research in this area later .

Summary ：

Performance optimization is a long-term process , There is no difference in system performance , The key is to look at the application scenarios . No system can cope with multi-user high concurrency at the beginning ,

All systems will experience a development process from small to large ,“ Yun Zhujia ” The same is true . If there is any mistake or unreasonable understanding in the above content , Please also make more corrections, great Xia .