Smooth live broadcast ｜ analysis of key technologies for live broadcast pain points

One 、 A sharp blade for upgrading live experience  ——「 Live broadcast 」

 

A good live experience has always been the key traffic password for live broadcast products to retain users .

 

The domestic live broadcasting industry has presented a red sea , The growth rate of users of various live broadcast products is gradually slowing down , The competition in the industry has entered a white hot stage , There is an urgent need to ensure the platform revenue through the refined operation of the live broadcast experience . For example, for major events 、 The head anchor optimizes the live experience .

 

For going to sea  APP  Even more so , Complex network environment 、 Model and equipment , As a result, live broadcast products face many experience problems , If the loading time of the first frame is too long 、 The picture is blurry 、 Caton et al , This has greatly affected the live broadcast viewing experience of users , Leading to the head anchor running away from home 、 Drop in revenue 、 Serious loss of users 、APP  Problems such as decreased activity , The causes of the above problems can usually be found from the network 、 equipment 、 Cost is summed up in three aspects .

 1、 Live experience is a roadblock

The Internet

• Complex network environment

• There are huge network differences between regions

• There are cross network situations

• CDN  Unstable quality

equipment

• There are many types of terminal equipment

• The proportion of low-end models in some regions is high 、 The problem is complicated

cost

• Use  RTC  The live broadcast effect is good , But the cost is high

• High definition of image quality increases the cost , It is also easy to cause a jam , How to determine the appropriate image quality is a headache

• The full link data monitoring system is complex

• How to better use the limited resources in high-yield places

 

2、「 Live broadcast 」 Full link upgrade one-stop live broadcast service

 

Facing the live broadcast experience ,2022  year  6  month  20  Japan , Global cloud communications service providers  ZEGO  That is, structure technology released 「 Live broadcast 」 Full link upgrade one-stop live broadcast service , Help the current live broadcasting industry to upgrade the live broadcasting experience ,

Create an ideal state of live broadcast experience optimization

 —— 

User centered , In seconds 、 hd 、 Take the optimal solution between fluency and other evaluation angles ; While refining and improving the user experience , Help enterprises save costs .

 

「 Live broadcast 」 The service scheme focuses on the time-consuming first frame 、 The quality of the picture is fuzzy 、 The whole link has been optimized for common live broadcast phenomena such as live broadcast Caton . At the same time  All-in-one SDK  The way of service , It's live 、 Real time audio and video 、AI  And so on , A smooth access  SDK, It can not only cover the live broadcast demand of the whole scene , It can also switch between real-time audio and video and live broadcast .

 

This article will focus on seconds 、 clarity 、 The three technical directions of fluency are analyzed , Take you to understand  ZEGO  That is to say, it constitutes the core technology optimization idea of science and technology in the Chang live broadcast service scheme .

 

Two 、 First frame optimization , Live broadcast seconds to enjoy the experience

The live broadcast industry usually pays more attention to the first screen loading time when the live broadcast is turned on , The fluency and clarity of sound and picture are indicators directly related to the user experience . It's like watching TV , Can't accept from CCTV  1  Switch to CCTV  2  On the channel , It takes a few seconds to see the picture , At the same time, I can't immerse myself in the blurred picture or even Caton's audio-visual world .

 

To reduce the response delay of channel switching , Enhance the user experience , Second open technology has become a rigid demand .

 

Let's first analyze how users click to enter the live broadcast room , The process that the user sees the picture and hears the sound ： First, assign an access point to the user , The user pulls the flow from this point , The process of allocating access points is called access scheduling ; Then the client needs to pull the stream indirectly with the access point ; If the access point does not have the flow, it also needs to introduce the flow from other servers to the access point , We call it back to the source ; When all this is done , The client can receive audio and video content , Play it .

 

Each of these steps will affect the secon experience .

1、 Customizable scheduling strategy

First, let's talk about access scheduling . The quality of access points directly affects the quality of streaming , It also directly affects the speed of construction . If the network between the client and the access point is poor , Like being  200ms  Of  RTT, So even if it can be in a  RTT  Built in connection and flow pulling , At least we need  200ms  Before you can see the first frame . If there is packet loss , Greater delays may be introduced . Conventional  CDN  Domain name resolution is used to specify the access point . To some extent, it solves the problems of nearby access and load balancing , However, it is still unable to achieve a more accurate and customized scheduling strategy .

 

ZEGO  To solve these problems , Self built dispatching system .

The system can customize the most appropriate scheduling scheme according to the customer's business model . That is to construct a self built unified access layer , Responsible for solving the access quality of the first kilometer of global users , It can ensure that the user can access to the access node with the best delay quality .

 

For example, we

It realizes the scheduling ability that can be accurate to the human level , In this way, when resources are limited , Give priority to ensuring the experience in the live broadcast room of popular anchors

. Simply put , Hot anchors get tagged , The marked hot anchor will get the best access resources in the whole network , The audience pulled the stream of popular anchors , You can also get the best access resources and bring a better viewing experience .

 

Another example is  

ZEGO  The most appropriate access point can be given according to the location of the source

： For example, the anchor pushes streaming in Shenzhen , If you simply follow the principle of nearby access , The audience will probably choose the access point in Guangzhou . In this way, the access point in Guangzhou needs to go back to Shenzhen , In this way, the distributed links become longer , Not only increase the cost , And it also introduces more delay and increases the loading time of the first screen .ZEGO  Because of the reference to the location of the source , The audience in Guangzhou can draw directly from Shenzhen , Without going back to the source . Of course , The premise here is that we believe that there is no link quality difference between Guangzhou and Shenzhen . The way of domain name resolution , Because the source information cannot be brought in , It is impossible to achieve such precise scheduling .

 

in addition , The quality of the link may be time-varying , There may be different optimal access points in different time periods , because  DNS  The reason for caching , Domain name resolution is also difficult to give time-varying scheduling results in time . All in all ,

ZEGO  The scheduling system takes into account the space , Time , Operator, , degree of heat , Location, etc , The optimal scheduling result is given

.

 

2、 Build connection and return to source

 

Then let's talk about Jianlian and Huiyuan . because  TCP  The existence of a protocol triple handshake , Build a  TCP  Links need at least  1.5 individual  RTT, Plus application layer data exchange , Users must be at least  2  individual  RTT  Then you can see the first screen . and

 ZEGO  By optimizing private protocols , Can achieve  0 RTT  Jianlian , At least in  1 RTT  Then you can show the first screen

.

 

in addition , An intuitive way to go back to the source is to go back to the source step by step , Can be seen as serial ：A  Go back to  B,B  Find that the stream resource does not exist and then go back to the source  C. This multi-level jump is common in transnational lines , Sometimes you need to  4-5  Hop to achieve a good transmission effect . However, the multi hop link will make the serial back to source method inefficient , The total time of returning to the source is the sum of all hops .

ZEGO  The parallel source returning method is adopted , namely  A  Back to the source  B,B  Back to the source  C  At the same time , In this way, the total time of returning to the source can be compressed to the maximum value in each hop

.

 

The above opens up the entire transmission link , The streaming end can receive audio and video data .

 

3、 Adaptive buffer technology of player

 

The last link that affects the seconds on is the player .

 

At present, many open source players in order to reduce the jam , You need to preset a playback buffer , The buffer is filled before decoding the rendering . If the buffer is set too small, it is easy to get stuck frequently , The buffer is too large , The buffer takes longer to fill , Directly affect the second opening experience .

 

and  

ZEGO  The player uses adaptive buffering technology

, During the playback , Real time adjust the size of the buffer according to the quality of the network to deal with the changes of the network . There is no problem that the buffer is filled before decoding and playing , Can be interpreted as ,

The moment the first frame is received, decoding and rendering have started

. Instead of filling the buffer before decoding the rendering , Suppose you set  1  Second buffer size , Even at the beginning of the pull flow  5  The transmission speed is times of the real-time code rate , It also takes to fill the buffer  200ms, The optimization here is remarkable .

 

After the seckai scheme goes online , The second opening rate of domestic market reached  99%, In all the second on actions ,85.07%  User in  500ms  Open inside , The second opening rate is relatively increased  14.5%, In Thailand and other regions with good network , Realization  96.8%  The second opening rate .

 

3、 ... and 、 Link quality , Ensure the clarity and smoothness of live broadcast

 

After talking about the second question , Let's talk about clarity and fluency . Naturally , Good link quality is the best way to ensure transmission quality , Good transmission quality enables users to use larger resolution and higher bit rate , To improve the video definition , Good transmission quality can also provide better guarantee of fluency .

 

But the reality may be disappointing , For example, when a dedicated line is not used , There is always a normal packet loss in cross-border links , For example, some countries from China to Europe always exist  200 ms RTT  and  30%  Packet loss around . Conventional  CDN  Use  TCP  transmitted , Even if more types of protocol access are realized in the first kilometer and the last kilometer of access , But it is still used between nodes  TCP  agreement .TCP  stay  200ms RTT  and  30%  Packet loss around can hardly be transmitted in real time , Even if it's  QUIC  Agreements don't work well .

 

ZEGO Through the self-developed transmission algorithm and its own protocol , The weak net performance has been improved very well .

 

Some people think that transmission is not difficult , Just multiple passes , Adding more redundancy can always achieve . in fact ,ZEGO  Aim for efficient transmission , The target of transmission is that each packet is received at the receiving end only once . If you receive a duplicate package , Duplicate packets are considered invalid transfers . Invalid transmission will bring more load to the network , Consume more bandwidth resources , Especially when the bandwidth is limited, the more invalid transmissions , The trafficability can only be obtained by reducing the bit rate of the bass video to a greater extent . therefore ,ZEGO  The goal is to ensure the smoothness and the invalid transmission cannot exceed  5%.

 

Another difficulty in transmission is , How to distinguish whether packet loss is caused by congestion or a normal packet loss . The difference between the two is that if congestion causes packet loss , Then reduce the bit rate properly , Packet loss will be alleviated or even disappear . In this case , It is necessary to actively reduce the bit rate , In order to avoid more packet loss and cause jamming . The normal packet loss is different , Packet loss does not change significantly due to the reduction of bit rate , let me put it another way , Packet loss is not “ you ” Caused by the . In this case, the better choice is not to reduce the bit rate . In this regard ,ZEGO  Our engineers worked hard , This is it.  ZEGO  It can be done  70%  It can still flow smoothly after dropping packets ,50%  While losing the package, I am smooth , The secret that the picture quality will not decrease significantly . in addition , combination  AI  Video restoration technology , Even if the link bandwidth is really only  100kbps, The block effect caused by insufficient coding rate can also be removed , Improve picture quality , Super resolution technology also provides guarantee for low code high definition .

 

in addition ,ZEGO  Around the world  200  many  IDC  Deployed edge media services , Provide users with the best access quality ; In some overseas areas, the network is complex , Quality parameters , We comprehensively use online big data guidance , And user side active detection technology to find the optimal coverage point , To optimize the access quality ; And in the process, we will monitor the change of quality , When the quality gets worse, we will enable dynamic detection to adjust the coverage points , So as to ensure the live broadcast quality .

Four 、 summary

The system architecture of live broadcast , It can flexibly and efficiently solve the problem of smooth full link , Support different business scenarios . One access , You can get  CDN  live broadcast 、CDN Plus、L3  And so on , And it can be targeted at regions 、 People and other dimensions , Through the console 、API  And so on , And smooth switching  RTC  Lian Mai .

 

All-in-One  One stop live broadcast service with full link upgrade , Second on 、 Fluent 、 Ultra HD image quality , One access , Enjoy live broadcast ！

 

More technical details are welcome to follow up ！