Tencent Cloud Video Live LVB Low Latency Technology: Farewell to "Spoiler" and "Slow Half-beat" Hard Core Dismantling
You must have encountered this maddening scene:
On New Year's Eve or watching the World Cup finals, you are in the official studio in front of the computer, and suddenly there is a cheer from the window or next-door neighbor: "The ball is in!" On the screen in front of you, the striker has just dribbling through midfield. At this moment, you are "spoilers" in the physical sense.
For example, in the e-commerce studio, the anchor shouted at the top of his voice: "the last three lucky bags, count down three two one, link up!" You poked the screen crazily, only to find that you couldn't grab it, because before you saw the picture, the three lucky bags were robbed by others five seconds ago.
This "slow half-shot" embarrassment is the number one killer of traditional live video--
Network latency
.
The traditional live broadcast technology, the picture from the host end to your mobile phone end, usually in the middle
3 to 5 seconds
, even up
More than 10 seconds
The lag. In order to break this deadlock, Tencent Cloud Live Video Broadcasting (LVB) has introduced hard-core technologies (such as the Fast Live WebRTC scheme) specifically for low-latency scenarios. Today, we will use the vernacular, not to talk about the dizzying communication formula, to take a deep look at how the Tengxunyun live broadcast "squeezed" the delay to the millisecond level.
Where did 1. traditional live broadcast delay the time?
To understand how powerful low-latency technology is, we have to be a "time auditor" to see where time is wasted when traditional live traffic travels on the public network.
The most commonly used protocol for traditional live streaming is called
RTMP (push flow)
and
HLS / FLV (Play)
. Their workflow is like "road freight":
The host side codes and slices: the host's camera takes pictures and the computer packs the pictures. If it is an HLS protocol, it must cut the video into small files (slices), each slice is usually 2 to 4 seconds. (Delayed 2-4 seconds)
Server Cache: Traffic is transmitted to CDN nodes in the cloud. In order to prevent network jitter from causing playback Catton, the server will deliberately save 2 to 3 slices in memory and send them down. (Another 4-8 seconds)
Player decoding: After your mobile phone gets the video, it is also afraid that the network is not good, and it will store a few seconds of data in the buffer of the player before it starts broadcasting. (1-3 seconds)
This set of processes came down, even in the smooth network environment, 5 seconds of time has slipped away unnoticed. This architecture is not inherently designed for "real-time interaction.
2. Tencent Cloud Fast Live (LEB): To set up a "quantum channel" for live broadcast"
In order to reduce the second-level delay to the millisecond level, Tengxun Cloud has carried out a "push-back" revolution on the traditional live broadcast architecture and launched a WebRTC-based
Live Live (Live Enhanced Broadcasting, LEB for short
)
.
It has done three things right at its core, directly suppressing the delay in death.
Within 1 second (usually around 500ms)
:
1. Protocol Dimension Reduction Strike: Switch from TCP to UDP (Fast Word)
Traditional live streaming is mostly based on the TCP protocol. TCP is an "obsessive-compulsive disorder". If a data packet is lost during transmission, all the subsequent packets must be stopped and queued until the lost packet is successfully retransmitted (I. e., queue head blocking). This is the culprit of delay surge in transnational or weak network environment.
Tengxunyun Fast Live has switched to the bottom layer
UDP protocol (more precisely WebRTC over UDP)
. UDP is like a generous courier, just rushing all the way forward. Did you lose your packet? It doesn't matter, the bottom layer cooperated with Tencent's self-research
PLC (Audio Loss Compensation)
and
FEC (Forward Error Correction)
Algorithm, through the algorithm directly to the lost pixel "guess" out to make up.
Don't wait, don't caton, run straight.
2. Throw away traditional slices: streaming, that is, on-the-fly.
Fast Live completely abandoned HLS's slicing mode of "saving enough seconds to pack once" and changed it
Pure streaming
. As long as the host side generates one frame of picture, the cloud will forward one frame and the player will render one frame. This is like upgrading from "taking a bus before leaving" to "taking a trolley bus as soon as one passenger comes", with no inventory turnover time at all.
3. Super Brain: Self-developed Node Network and Intelligent Routing
Tengxunyun has super infrastructure in the world (EdgeOne edge security acceleration platform, etc.). The traffic of fast live broadcast is not blindly running on the public network, but entering Tencent's private high-speed network.
There are AI intelligent routes inside this network, which can sense in real time which fiber is stuck in traffic, which backbone node is losing packets, and in
Millisecond level
Complete path switching. At the same time, it will dynamically adjust the contracting strategy according to the current WiFi signal strength of the user's mobile phone.
3. Scene: What Industries Has Low Latency Live Changed?
A lot of people will say, "I'll just watch a party, can I lose a piece of meat by slowing down for 5 seconds?" Indeed, ordinary TV watching does not require low delay. But in the following real money business scenarios, low latency technology is the lifeline of the enterprise:
Scene 1: E-commerce Live with Goods (Real Dignity of Seconds)
In the traditional 5-second delay studio, the anchor shouted "start robbing". in fact, it has been 5 seconds since the first batch of users who saw the picture, while users with slow network speed may not see it until 10 seconds later. This not only caused absolute unfairness, but also greatly hit the user's sense of participation.
Low latency enabling: after accessing tengxunyun fast live broadcast, the delay of the whole network is reduced to about 500ms. The anchor's voice just fell, the users of the whole network saw the shopping cart pop-up window almost at the same moment, the adrenaline of ten thousand people killing on the same screen was instantly filled, and the conversion rate soared directly.
Scene 2: Online Education and Large Classes (Teachers Don't Have to Wait for Questions)
In the online large class, the teacher wrote a question on the whiteboard and asked, "Boys and girls, do you choose A or B for this question?"
Pain point: under the traditional live broadcast, the teacher has to wait awkwardly in front of the camera for 10 seconds after asking the question before he can see the feedback from the students on the public screen, and the teaching rhythm is sparse.
Low-delay empowerment: the delay is shortened to one sentence, teacher-student interaction is like face-to-face communication, and large classes can also give a sense of interaction in exquisite small classes.
Scene 3: Sporting Events and Pan-Entertainment (Farewell Spoilers and Even McCarton)
Whether it's the live broadcast of the ball match or the double anchors in the live broadcast of the show play PK (Lian Mai) across the room from a distance.
Low Delay Enabling: Fast live broadcast can ensure that you are almost synchronized with the whistle on the scene, completely eliminating the pain of being spoiled by your next-door neighbor in advance. During the war between Lian and Mai, the dialogue between the two anchors was smooth and smooth, and there would be no embarrassing fault of "you say it, I'm in a daze.
4. architects must see: is there any side effect of accessing fast live (low latency)?
As a rational technical person, we must admit:
There is no free lunch in the world, technology is always doing balance (Trade-off).
Since the fast live broadcast is so good, what is the price? What should we pay attention to when selecting the type?
The cost of bandwidth has increased slightly: in order to combat packet loss under UDP network, fast live will use FEC (forward error correction) and other technologies to send some additional redundant "recovery packets". As a result, under the same picture quality, the bandwidth traffic consumed by fast live broadcast is usually slightly higher than that of traditional FLV by about 10%. Businesses need to include this cost when making a budget.
Client compatibility (H5 web side to avoid pits): Although mainstream modern browsers (Chrome, Safari, Edge) now perfectly support WebRTC, WebRTC compatibility occasionally steps on the thunder in some old Android models built-in browsers or in some specific App WebView. Tencent Cloud's downgrade strategy: Tencent Cloud LVB SDK provides a very smart automatic downgrade mechanism. If it detects that the user's device does not support extremely fast live streaming, it will automatically and seamlessly switch back to the standard low-latency HTTP-FLV or the standard HLS protocol. Ensure that "fast is fast, not fast and never black screen".
5. Summary
Network delay is essentially a tug-of-war between human technology and physical speed of light and public network congestion.
Tencent Cloud Video Live LVB successfully grafted the real-time communication technology originally belonging to professional audio and video conferencing into the general live broadcast field with millions of concurrency through the underlying fast live broadcast (WebRTC) technology. It has flattened the "second mountain", which the industry used to think was insurmountable, to the "millisecond plain".
Today, when the dividends of the live broadcast industry are peaking and moving towards refined operation, whoever has a more silky technical experience and a more real-time interaction can dig out more conversion rates in the rapidly changing market. And Tencent Cloud LVB's low-latency technology is undoubtedly that.
To help enterprises break the gap between time and space "breaking the blade".