Qos（Quality of Service）

QOS（ namely Quality of Service, Service quality ） It mainly refers to the degree to which the service meets the users in the network environment , In the context of video service, it can also be regarded as Quality of Streaming, That is, the quality of streaming media services . Usually ,QOS It can be expressed by a series of indicators , Such as transmission speed 、 response time 、 Sending order 、 Accuracy, etc . In terms of video services ,QOS It consists of a number of conventional technical indicators , Include Playback success rate 、 Error rate 、Re-buffer（ Carton ） Times and times 、 Starting time 、 Fast forward response time 、 Video rate 、 Delay etc. .

The passage of the QOS Indicators can be roughly divided into two categories ：

One is used to measure What is the probability that users will get the service , Such as playback success rate and error rate ;

The other describes The level of service that users get , Such as Number of catons 、 Time 、 Starting time 、 Fast forward time 、 Video rate and Delay

Here is QOS Index definition （ The picture is from Conviva Official website ）

• Playback success rate ： Describes the success rate of startup when users try to play video , You can divide the number of successful starts by the total number of user attempts , Common in Back end video failure The circumstances of
• Playback error rate ： An error that occurs when at least a single video or audio frame is played during playback , Possible causes include The player crashed 、 Hardware off 、 The network is disconnected etc. , User intervention is required to resume playback
• Play the opening video 、 Failure caused by video or interactive advertising ： These may be caused by third-party services , Impact on user experience , It should also be monitored and debugged for improvement
• Video Caton ： The player cannot get the video clips transmitted by streaming media immediately and needs to wait for downloading . Caton can happen in a short time , It may also last for a long time . The Caton indicator includes both the number of catons per unit time and the dimension of cumulative Caton time , A common way to optimize the Caton time is utilize CDN and Rate adaptive algorithm .
  • A special case of video jam is the jam at the beginning of playback , It usually calculates the length of time from the user clicking to play until the first frame is displayed on the screen , Because it takes time to get the first available video clips , Include Background service prepares resources 、 Download the video start clip 、 Initialize hardware and software etc. .
  • The start time can be optimized by pre placing the start video clip in CDN Edge node of , Reduce the initial bit rate , Increase player initialization parallelism , Pre established network connection Methods such as . Besides , The player can also Insert title animation , Keep playing the video clip before fast forward Reduce the subjective waiting time of users until the video frame after fast forward is ready .
• Average bitrate ： Used to reflect the clarity of the video , Different encoders , Or even different coding parameters will lead to the same level of video bit rate difference , Therefore, this index is mainly used to evaluate the quality of streaming media services , Does not fully represent the viewing experience
• Program delay time ： The calculation standard is the interval between the time that the program should be broadcast and the time that is actually broadcast on the screen . The software processing speed that causes delay 、 The network transmission speed is beyond , Encoder , Source server and CDN The cache queue brought by the server , The decoder and rendering hardware in the player will introduce delays of different sizes .

When defining the key QOS After indicators and priorities , Video services can be provided in traditional ways （ Such as local environment 、 Test environment test method ） Perform initial verification , Use as Evalvid Such a video quality evaluation tool , coordination NetworkSimulator Class network simulation tool , It can be used for small-scale test , Provide packet loss rate 、 Delay 、 Jitter and other low-level indicators .

In real online video services , Most users are located in different geographical locations 、 Types of equipment 、 Software version 、 Problems caused by the running environment are difficult to reproduce in the development environment . Online video services are usually implemented on various client platforms QOS Condition monitoring , adopt SDK Send back , The back-end service performs conversion and statistics .

When the number of users reaches a certain level , Need big data technology to distinguish between real-time and batch data , And pretreatment before storage .QOS The data will be applied to the chart presentation after being integrated by the background service 、 Statistical report 、 Analysis optimization 、 Monitoring alarm and other purposes , It's the product 、 Development 、 Operation and maintenance 、 Data analysis is the foundation on which the team relies .

For better analysis of specific problems , Collect all information about a user's playback process and present it in time sequence , Can effectively help understand the causal relationship , The information will include User behavior 、 execution time 、 Download timing 、 Rate switching record 、 Wrong type 、CDN Node location 、 Server log Even some Intermediate results of calculation , It will be possible to effectively infer, for example, the reason why the playback starts slowly or how a certain jam occurs .

adopt QOS Data and extended information , Developers are free to optimize , The optimization aspects that can be considered include architecture design 、 Encoding selection 、 Streaming protocol 、 Adaptive algorithm 、 Connection and Caton logic 、 Client software design .