当前位置：网站首页>Teach you how to realize an efficient im long connection adaptive heartbeat keeping mechanism

Teach you how to realize an efficient im long connection adaptive heartbeat keeping mechanism

2022-07-21 22:53:00 【hellojackjiang2011】

The author of this article “Carson”, Now working in Tencent , The original title is “ High efficiency and long connection ： Hand in hand to teach you to realize the adaptive heartbeat preservation mechanism ”, There are many revisions and changes .

1、 introduction

When you want to achieve IM Instant messaging chat 、 When high real-time requirements such as message push , We usually choose the communication mode of long connection .

And when the long connection mode is really realized , There will be a lot of technical problems , For example, the most common problem of keeping long connections alive .

today , I will pass this article , Hand in hand to teach you to realize a set of adaptive heartbeat preservation mechanism , Thus, it can efficiently and stably maintain such as IM Long connections for such needs as chat .

Exchange of learning ：

- Mobile IM Introduction to development ：《 Beginner level one is enough ： Develop mobile from scratch IM》
- Open source IM The framework source code ：https://github.com/JackJiang2011/MobileIMSDK

（ This article is published synchronously in ：http://www.52im.net/thread-3908-1-1.html）

2、 Related articles

3、 What is long connection

Know the long connection ：

The main function of long-term connection is to keep both sides connected for a long time , thus ：

1） Speed up communication ;
2） Ensure real-time ;
3） Avoid the waste of channel resources and network resources caused by repeated connections in a short time .

The difference between long connection and short connection ：

PS： about IM For developers like this , Usually people put HTTP The agreement says “ Short connection ”、 Based directly on TCP、UDP or WebSocket Of socket be called “ A long connection ”.

4、 Cause of long connection disconnection

4.1 Basic concepts

From the last section we can see that , With long connections , All communications between the parties are based on 1 The length of the bar is connected to （ such as 1 Time TCP Connect ）. therefore , A long connection requires a continuous connection between both parties to enable both parties to communicate continuously .

However , The reality is that , The connection will be broken for a long time .

The main reasons for these disconnections are ：

1） The process of the long connection is killed （ This mainly refers to the mobile terminal ）;
2）NAT Overtime ;
3） Network state changes ;
4） Other irresistible factors （ The state of the network is poor 、DHCP And so on ）.

below , I will analyze each reason .

4.2 make a concrete analysis

1） reason 1： Process killed

When the process is killed , The long connection will also be broken . The process was killed in Andriod End is the most common problem , Limited to space, I will not start this topic here , If you are interested, you can read this ：《Android P The official edition is coming ： Backstage application to keep alive 、 The real nightmare of news push 》.

2） reason 2：NAT Overtime （ Focus on ）

NAT The overtime phenomenon is as follows ：

Of each operator and region NAT The timeout is as follows ：

PS： The above data comes from the wechat team 《 Mobile IM practice ： Realization Android Version of wechat's intelligent heartbeat mechanism 》 One article , With 4G、5G The popularity of , These data may have changed , Please refer to the actual test results .

Particular attention ： Exclude other external causes （ Network switching 、NAT Overtime 、 For human reasons ）,TCP A long connection is in a situation where both sides are constantly disconnected , It doesn't automatically interrupt （ That is, you don't need a heartbeat package to maintain , You can verify that ： Give Way 2 A computer is connected to 1 individual Wifi, among 1 Taiwan as a server , another 1 Taiwan does client connection server （ No settings KeepAlive）. As long as the computer 、 The router is constantly losing power , that ,2 The long connection of a computer is not automatically interrupted ）.

Jack Jiang notes ： The above statement may not be accurate , Readers with new interests can read in detail 《 Unplug the network cable and plug it in ,TCP Is the connection still there ？ You can understand... In a word ！》.

3） reason 3： Network state changes

When the mobile client network state changes （ Like the mobile network & Wifi Switch 、 To break off 、 Reconnection ）, It also breaks long connections .

4） reason 4： Other irresistible factors

For example, the network status is poor 、DHCP And so on , Will make the long connection happen Accidental disconnection .DHCP The lease term of ： about Android System , DHCP When it comes to the lease term, it will not renew the contract voluntarily （ Continue to use expired IP）, The connection is long, resulting in disconnection .

5、 Efficient solution to maintain long connections

5.1 Basic introduction

After understanding the reason why the long connection is disconnected , For these reasons , Here is my solution to efficiently maintain long connections （ As shown in the figure below ）.

So , To effectively maintain long connections , It needs to be done ：

Put it simply , The key to maintaining a long connection efficiently is ：

1） Keep alive ： Try not to disconnect when you are connected ;
2） Reconnection ： After the connection is broken, you should be able to continue to reconnect .

5.2 Specific measures

1） measures 1： Process keeping alive

The overall summary is as follows ：

PS： About Android Keep the process alive , This topic is very popular , If you are interested, you can read the following article in detail ：

2） measures 2： Heart beating mechanism

This is the focus of this article , The next section begins with a detailed analysis of

3） measures 3： Disconnection and reconnection mechanism

The principle is ： Detect the change of network state and judge the effectiveness of connection in time .

Concrete realization ： This is actually a complete set of logic with the heartbeat preservation mechanism , So the following will be explained in the heart beating mechanism .

6、 Brief introduction of heart beating mechanism

The overall introduction of heart beating mechanism is as follows ：

however , Many people easily confuse the heartbeat mechanism with the traditional HTTP The polling mechanism is confused .

Here is the difference between the two ：

7、 Main stream IM Analysis and comparison of heartbeat mechanism

To the country 、 The movement of internal and external mainstream IM product （WhatsApp、Line、 WeChat ） Simple analysis and comparison of heartbeat mechanism .

Please see the following picture for details ：

PS： The above data comes from the data shared by wechat team 《 Mobile IM practice ：WhatsApp、Line、 Heartbeat strategy analysis of wechat 》 One article .

8、 Overall design of heartbeat preservation mechanism scheme

below , I will follow the mainstream heartbeat mechanism on the market , A scheme of heartbeat mechanism is designed .

The basic flow of heartbeat mechanism scheme ：

The main consideration for the scheme design of heartbeat mechanism is ：

1） To ensure the real-time performance of messages ;
2） Consider the resources that consume the equipment （ The network traffic 、 Electric quantity 、CPU wait ）.

As can be seen from the above figure , The key to the design of heartbeat mechanism is ：

1） The specification of the heartbeat bag （ Content & size ）;
2） The interval between heartbeat sending ;
3） Disconnection and reconnection mechanism （ The core = how Judging the validity of long connections ）.

In the following scheme design , Will aim at this 3 Give a detailed solution to the problem .

9、 Detailed design of heartbeat mechanism scheme

9.1 The specification of the heartbeat bag

In order to reduce traffic and improve transmission efficiency , Need to streamline the design of the heartbeat pack .

It mainly starts with the content and size of the heartbeat package , The design principles are as follows ：

design scheme ：

Heartbeat bag = 1 Carry a small amount of information & Size in 10 Packets in bytes

9.2 The interval between heartbeat sending

in order to prevent NAT Timeout and reduce the consumption of device resources （ The network traffic 、 Electric quantity 、CPU wait ）, The interval of heartbeat transmission is the focus of the whole heartbeat mechanism scheme design .

The design principle of heartbeat interval is as follows ：

9.3 The most commonly used heartbeat interval scheme

commonly , The most direct and commonly used heartbeat transmission interval setting scheme mostly adopts ：“ Every estimate x Minutes to send heartbeat packets 1 Time ”. among ,x ＜5 minutes （ Integrated mainstream mobile IM product , It is suggested that x= 4 minute ）.

however , There are some problems with this scheme ：

PS： The specific implementation of the scheme of fixed heartbeat interval , You can refer to ：

9.4 Adaptive heartbeat interval scheme

below , I will explain in detail the design scheme of adaptive heartbeat interval .

Basic logic ：

The solution needs to be solved 2 A core question .

1） How to adaptively calculate the heartbeat interval So that the heartbeat interval near At present NAT Timeout time ？

answer ： Increasing heart rate interval for heart rate response test , Until the heartbeat fails 5 Next time , You can find the closest to At present NAT Time out of the heartbeat interval .

Please see the following picture for details ：

notes ： Only when the heartbeat interval near NAT Timeout time when , To maximize the balance Long connection is not interrupted & The lowest consumption of equipment resources .

2） How to detect In the current network environment NAT Timeout time There is a change ？

answer ： Sending heartbeat packet successfully The maximum interval time of （ That is, the closest to NAT The heartbeat interval of the timeout period ） fail in send 5 Next time , Then judge the current network environment NAT Timeout time There is a change .

Please see the following picture for details ：

notes ： In the detection of NAT Timeout time After the change , Recalculate the heartbeat interval adaptively So that the heartbeat interval near NAT Timeout time

To sum up ： Overall planning above 2 A core question , The design scheme of self adaptive heartbeat interval is summarized in the figure below ：

PS： Design and implementation of adaptive heartbeat mechanism , You can refer to ：

10、 Implementation of disconnection and reconnection mechanism

Technically ： Keeping the heartbeat alive for long connections depends on the heartbeat mechanism , When the heartbeat mechanism works , Timely start the disconnection and reconnection mechanism , Only under the joint action of heartbeat mechanism and disconnection reconnection mechanism can the real heartbeat be kept alive . But to make the logic clearer , I will explain the disconnection reconnection mechanism and the heartbeat mechanism in a separate section . This section is about the fragment line reconnection mechanism .

The core of the mechanism is ： How to judge the validity of long connections . namely ： Under what circumstances is a long connection considered broken ？

1） Design principles ：

The basic logic is ： Criteria for judging whether a long connection is valid = Whether the server returns a heartbeat response .

Here, we need to distinguish the long connection “ Survive & It works “ The difference between States ：

2） The specific plan ：

Realize the idea ： By counting , If continuous 5 After sending heartbeat times , No heartbeat response from servers , The long connection is considered invalid .

Judgment process ：

3） The scheme circulated on the Internet ：

There are some schemes circulating on the Internet to judge whether the long connection is effective , The details are as follows ：

thus , The mechanism of heart beating is explained .

11、 Program summary

It is necessary to summarize the heartbeat mechanism and disconnection reconnection mechanism I shared in the last two sections , These two mechanisms constitute the long connection heartbeat keep alive complete logic in this paper .

design scheme ：

Process design ：

notes ： identification “ gray ” Please refer to the above description for the judgment process of .

12、 Further optimize and improve the heartbeat preservation scheme

12.1 Basic information

The solutions in the previous two sections will still have technical defects , The connection is long, resulting in disconnection （ such as ： The long connection itself is not available （ It doesn't matter how many times you have to do it again ））.

The following will optimize and improve the above scheme , To ensure that The client and the server are still communicating .

The main optimization point is ：

1） Ensure the effectiveness and stability of the current network before starting a long connection ;
2） Timing of adaptive calculation of heartbeat packet interval time .

12.2 Ensure the effectiveness and stability of the network before starting a long connection

Problem description ：

Solution ：

Add to the main process of the original heartbeat preservation mechanism ：

12.3 Timing of adaptive calculation of heartbeat packet interval time

Problem description ：

The project design ：

Add to the main process of the original heartbeat preservation mechanism ：

12.4 To sum up

13、 Extra thinking ：TCP Provided by the agreement KeepAlive Whether the mechanism can replace the heartbeat mechanism ？

Many people think that ,TCP The agreement itself has KeepAlive Mechanism , Why based on its communication links , Additional heartbeat keeping mechanism still needs to be implemented in the application layer ？

The conclusion is that ： There is no substitute for ;
as a result of ：TCP KeepAlive The function of the mechanism is to detect whether there is a connection （ Life and death ）, But we can't detect if the connection is valid .

notes ：“ The connection works ” The definition of = Both parties have the ability to send & The ability to receive messages .

First look at it. KeepAlive What is the mechanism ：

KeepAlive The specific reasons why the mechanism cannot replace the heartbeat mechanism are as follows ：

Particular attention ：

1）KeepAlive Mechanism is just a passive mechanism at the bottom of the operating system , Should not be used by the upper application layer ;
2） When the system is shut down, a KeepAlive When the mechanism checks for dead connections , It won't actively notify the upper application , Only by calling the corresponding IO Found in the return value of the operation .

To sum up, it is ：KeepAlive The mechanism can't replace the heartbeat mechanism , It needs to be in the application layer Implement heartbeat mechanism to detect the validity of long connection , Thus maintaining long connections efficiently .

Jack Jiang notes ： About TCP Of itself KeepAlive Mechanism , May read in detail ：