Interviewer: why does database connection consume resources? Where are the resources consumed?

*

Developing applications for a long time , Always want to get to the bottom , Especially for some questions with public answers . Everyone can explain , But in the final analysis , Can't explain clearly . Everything has a reason , And it is the most intuitive to explain the problem with numbers .

*

This article mainly wants to explore the details of connecting to the database , Especially in Web In the application, the database should be used to connect to the pool , Avoid re establishing a connection every time you send a request . For this question , The answers are consistent , Establishing a database connection is time-consuming , But how much time does it take , What are the time-consuming aspects ？

This paper is connected with MySQL Database, for example , because MySQL The database is open source , Its communication protocol is open , So we can analyze the whole process of establishing connection in detail .

*

In this paper , The analysis of resource consumption mainly focuses on the network , Of course , Resources also include memory 、CPU Equal computing resources , The programming language used is Java, However, it is not ruled out that programming languages will also have a certain impact .

*

First, let's take a look at the connection to the database Java Code , as follows ：

Class.forName("com.mysql.jdbc.Driver");

String name = "xttblog2";
String password = "123456";
String url = "jdbc:mysql://172.16.100.131:3306/xttblog2";
Connection conn = DriverManager.getConnection(url, name, password);
//  Then the program terminates , The connection was forced to close 

And then through 「Wireshark」  Analyze the whole connection establishment process , as follows ：

Wireshark Grab the bag

In the connection process shown in the figure above , It can be seen that MySQL The communication protocol is based on TCP Transmission protocol , And the protocol is binary , It's not like HTTP Text protocol for , The process of establishing connection is as follows ：

• The first 1 Step ： establish TCP Connect , Through three handshakes ;

• The first 2 Step ： The server sends it to the client 「 Handshake messages 」 , The client responds to the handshake message ;

• The first 3 Step ： client 「 Send authentication package 」 , For user authentication , After successful verification , Server return OK Respond to , Then start executing the command ;

After successful user authentication , Some connection variables will be set , For example, the character set 、 Whether to automatically commit transactions, etc , There will be multiple data interactions . After completing these steps , Will perform real data query, update and other operations .

In the test of this paper , It only took 5 Line of code to establish a connection , But no operation is performed through this connection , So after the program is executed , The connection is not through Connection.close() closed , But because the program is finished , Cause the process to terminate , Cause the connection with the database to close abnormally , So in the end TCP Of RST message . In this simplest code , No additional connection properties have been set , Therefore, the time spent on setting properties can be considered to be the least （ Actually , We didn't set any properties , But the driver still sets the character set 、 Automatic transaction submission, etc , This depends on the specific driver implementation ）, So the whole connection takes the least time . But statistics show that , Excluding the last TCP Of RST  When the message （ Because the message does not require the server to return any response ）, However, there is still a round trip between the client and the server 「7」  Time ,「 That is, complete a connection , It can be said that , Data needs at least a round trip between the client and the server 7 Time 」 , In terms of time , From the beginning TCP Three handshakes of , Until the final connection is forcibly disconnected （ Not including the last RST message ）, Total cost ：

10.416042 - 10.190799 = 0.225243s = **225.243ms**！！！

It means , Establishing a database connection requires 225ms, And this can still be regarded as the least , Of course 「 The time spent may be affected by network conditions 、 Database server performance and application code efficiency 」 , But here is just the simplest example , It's enough to say ！

Because the above program terminated abnormally , But in a normal application , The connection is usually closed through Connection.close() Accomplished , The code is as follows ：

Class.forName("com.mysql.jdbc.Driver");

String name = "shine_user";
String password = "123";
String url = "jdbc:mysql://172.16.100.131:3306/clever_mg_test";
Connection conn = DriverManager.getConnection(url, name, password);
conn.close();

In this case , Things have changed , It is mainly reflected in the disconnection from the database , Here's the picture ：

Network packet capture

• The first 1 Step ： At this time MySQL Communication protocol phase , The client sends a request to close the connection , And you don't have to wait for the response from the server ;

• The first 2 Step ：TCP disconnect ,4 Complete the connection and disconnection with one wave ;

Here is the complete completion from the establishment to the closure of the database connection , The whole process took a long time ：

747.284311 - 747.100954 = 0.183357s = 183.357ms

There may also be the impact of network conditions , More than the above 225ms Less , But it has almost reached 200ms The level of .

So here comes the question , Imagine the scene , For a day job 2 For WAN's website , Assume that each user will only send 5 A request , So one day is 10 Million requests , For establishing database connection , Let's calculate conservatively as 150ms Okay , How much time do you spend establishing database connections in a day （ It does not include performing query and update operations ）：

100000 * 150ms = 15000000ms = 15000s = 250min = 4.17h

In other words, the time spent on establishing database connection every day has reached 「4 Hours 」 , So database connection pool is a must , And when the day's activity increases , Using database connection pool alone can't completely guarantee the normal operation of your service , Other solutions need to be considered ：

• cache

• SQL Precompilation

• Load balancing

• ……

Of course, this is not the main content of this article ,「 The core idea of this paper is only one , Database connection is really time-consuming , So don't connect frequently 」

  Please look forward to my next article , thank you .

more java Advanced information , Interview information , Official account