当前位置:网站首页>New experience of mlvb cloud live broadcast: millisecond low latency live broadcast solution (with live broadcast performance comparison)
New experience of mlvb cloud live broadcast: millisecond low latency live broadcast solution (with live broadcast performance comparison)
2022-07-27 01:09:00 【Zhimi Technology】
With the continuous expansion of mobile live broadcast scenes , From the beginning of the traditional scene , For example, outdoor exploration 、 Online online classes have gradually appeared in selfie talent or game live broadcast 、 New types of scenes such as interactive answer or live broadcast of sports competitions . meanwhile , These new scenarios have a common requirement for mobile live broadcast, that is, lower latency .
Based on this demand , Extended based on Tencent cloud MLVB Mobile live broadcast low delay live broadcast scheme —— Live broadcast , The full name is Live Event Broadcasting perhaps LEB. Today, Zhimi technology will analyze the advantages and specific performance of this low latency live broadcast scheme .
Advantages of fast live broadcast
- Millisecond delay
because MLVB The live broadcast plug-in adopts UDP Communication protocol of , Therefore, it can still be implemented in the case of high concurrency 1 Delay within seconds , That is, the delay reaches the millisecond level . Compared with ordinary mobile live broadcast, it may be as high as 5 Second is more fluent . At the same time, it can also take into account the situation of second opening and Caton .
- Strong compatibility
Fast live broadcast products have low latency , It is still compatible with the push-pull stream in the standard mobile live broadcast 、 transcoding 、 Recording 、 Screenshot 、 Chi Huang 、 Play and other full functions . Therefore, engineers can develop from ordinary MLVB Move live broadcast to LEB Fast live Application .
- High coverage of server nodes
Because the product is backed by Tencent cloud , Therefore, fast live broadcast has also distributed server acceleration nodes all over the world , Currently, there are 2 Thousands of nodes , covers 25 A country . meanwhile , The supported bandwidth also exceeds 100T
- Quick start
As a mobile live broadcast MLVB A product branch of , Fast live also uses standard protocols , Therefore, connect with other platforms and APP It's also relatively simple . In Google browser and Apple's own Safari Can be played directly in .
- Network fluctuation resistance
It is also due to fast live broadcast and MLVB They are all under Tencent cloud product line , Therefore, high-quality cloud servers can ensure the quality of video streams in a relatively weak network environment .
- Web Low delay
at present CDN Live broadcast only supports HLS Format stream , But the playback delay of this format is also up to several seconds . And fast live broadcast can also support web side playback , And there is only a delay of hundreds of milliseconds .
- Seamless multi rate switching
Seamlessly switch transcoding streams with different bit rates , There will be no interruption or jump in the switching process , Realize the smooth transition between appearance and hearing .
- Adaptive rate control
Adaptively adjust and switch different code streams according to the network bandwidth , Ensure smooth playback experience when different network conditions change
Effect comparison : Normal network environment vs Weak network environment
- Test scenarios
The host side uses RTMP Push flow , The audience plays separately FLV And fast live streaming , Statistics of Caton rate and other indicators . The anchor is a lossless network , Different weak networks are set on the audience side for testing . The main test indicators are frame rate and Caton rate .
- Streaming parameter configuration
The resolution of the :1080P
Bit rate :1800 kbp
Frame rate :15 frame / second
- Comparison of key indicators of live broadcast
Video frame rate

- Video card frame rate

- Audio card frame rate

Parameter description
Video card frame rate : The video rendering interval is greater than 500 Milliseconds are regarded as stuck , The total of all the jam times divided by the total playback time is the jam rate
Audio card frame rate : The audio playback interval is greater than 500 Milliseconds are regarded as stuck , The total of all the jam times divided by the total playback time is the jam rate
Video frame rate : Video frames per second
Data sources : live broadcast SDK Advantages of fast live broadcast
边栏推荐
- Flink based real-time computing Demo - Data Analysis on user behavior
- Redisson working principle - source code analysis
- forward和redirect的区别
- Flink 1.15本地集群部署Standalone模式(独立集群模式)
- Golang implements AES with five encryption mode functions, encrypt encryption and decryption string output
- VMware Workstation 虚拟机启动就直接蓝屏重启问题解决
- MySQL索引优化:哪些情况下需要建立索引(适合构建索引的几种情况)
- flink1.11 sql本地运行demo & 本地webUI可视解决
- FlinkSql多表(三表) join/interval join
- Spark on yarn's job submission process
猜你喜欢

Simple explanation of database table connection

Flink based real-time project: user behavior analysis (II: real-time traffic statistics)

Flink checkpoint源码理解

Android -- basic usage of litepal database framework

腾讯云直播插件MLVB如何借助这些优势成为主播直播推拉流的神助攻?

李宏毅机器学习(2021版)_P5-6:小梯度处理

Flink 1.15实现 Sql 脚本从savepointh恢复数据

深度学习报告(3)

MySQL索引优化:哪些情况下需要建立索引(适合构建索引的几种情况)

Flink 滑动窗口理解&具体业务场景介绍
随机推荐
MySQL第一篇
Valueerror: the device should not be 'GPU', since paddepaddle is not compiled with CUDA
MySQL index optimization: under what circumstances do you need to build an index (several situations suitable for building an index)
Status management in Flink
SQL学习(3)——表的复杂查询与函数操作
基于Flink实时项目:用户行为分析(三:网站总浏览量统计(PV))
Spark源码学习——Data Serialization
2022.7.18DAY608
C # conversion of basic data types for entry
Write the changed data in MySQL to Kafka through flinkcdc (datastream mode)
onSaveInstanceState和onRestoreInstanceState方法的调用
Flink based real-time project: user behavior analysis (II: real-time traffic statistics)
Uni-app 小程序 App 的广告变现之路:Banner 信息流广告
Flink1.11 intervalJoin watermark生成,状态清理机制源码理解&Demo分析
MLVB 云直播新体验:毫秒级低延迟直播解决方案(附直播性能对比)
基于Flink实时计算Demo—关于用户行为的数据分析
李宏毅机器学习(2017版)_P6-8:梯度下降
Real time calculation demo based on Flink: user behavior analysis (IV: how many different users have visited the website (UV) in a period of time)
Spark源码学习——Memory Tuning(内存调优)
Redisson working principle - source code analysis