Quickly generate thumbnails, Serverless supports event broadcast and locks down the highlights of the Winter Olympics

Author | Xiliu, Xiao Jiang
​
"The Beijing Winter Olympics became the most watched Winter Olympics in history on the fourth day of the competition, and its total broadcast content production will reach 6,000 hours, surpassing the 5,400 hours of the Pyeongchang Winter Olympics. Follow Beijing Winter Olympics With more crowds than ever, the Beijing Winter Olympics is expected to attract more than 2 billion viewers worldwide." This was released by Ianis Exarks, CEO of Olympic Broadcasting Service (OBS) on February 10. a set of data. There is no doubt that the Beijing Winter Olympics has been the "top stream" in the world network for the past month.
​

The enthusiasm of the global audience for watching the Beijing Winter Olympics has made the broadcast video of the event attract much attention. By browsing event highlight thumbnails, viewers can quickly find what they are interested in among the many broadcast videos. For broadcasters, how to simply and quickly process the live video of massive events in real time and generate thumbnails at different points in time is a difficult problem.

Serverless supports event broadcast to lock down the highlights of the Winter Olympics

The video thumbnail is the thumbnail of a frame of animation in the video, which is equivalent to converting the cover of the video or the content of one of the frames into a picture file. The selected video thumbnail can highlight the bright spots in the video and quickly capture it. Keeping users’ attention and improving the click-to-play rate, the thumbnails of competition videos are somewhat different from those of ordinary videos, and it is necessary to capture a large number of videos more quickly and in real time.
​

Customers supported by Function Compute need to perform real-time processing on the massive live video of the Beijing Winter Olympics, and take screenshots at different time points in the live video (the technical term "video image capture"). The customer's requirements are as follows:

• Need timed events to trigger screen capture operations
• The customer only wants to concentrate on the specific logic development, the core video screenshot logic code is not large, and the development needs to be completed quickly
• Customers don't want to manage and deploy virtual/physical machines, they want the simpler the better
• Customers need this core business logic to be flexible and highly available, free of operation and maintenance

By using Alibaba Cloud Function Compute (FC), users can develop the core business code of video screenshots, with low development cost and one-click deployment, to obtain a flexible, high-availability and O&M-free live video screenshot service, helping to realize the highlights of the Winter Olympics video. Crawl in real time.
​

picture only shows video thumbnail effect
​

In fact, as early as the 2021 Tokyo Olympics, some broadcast companies have used Alibaba Cloud Function Computing (FC) video thumbnail service to perfectly support live events to lock in the highlights in real time. During the 15-day period of the Olympic Games, tens of millions of function calls occurred, and FC ensured that the service is elastic and highly available.

Stable response to explosive peak traffic, real-time multimedia processing

The public cloud serverless architecture has the advantage of coping with explosive peak traffic. Function Compute is a serverless computing platform provided by Alibaba Cloud. Function Compute can dynamically allocate execution environments according to the amount of requests and schedule computing resources in milliseconds to ensure stable latency when the load is high and high resource utilization when the load is low. , and only pay for the computing resources your code uses when it runs. Function Compute can also be seamlessly integrated with object storage services, which can easily process images stored in object storage in real time.

1. Function Compute video live video clipping service

​

In the business scenario of Yidui App, the dating and dating product of Beijing Milian Technology Co., Ltd., live video is the most important link. Based on the skeleton of live video, it can be integrated into various innovative business models such as online matchmakers. content security puts forward extremely high requirements. Whether it is to intelligently analyze the live video content through AI technology, or to respond to regulatory requirements, it is necessary to cut frames of the video according to a fixed frequency after each live video stream starts, and generate the cut frames through a unified audit service. images are processed.
​

In this requirement, the frame-cutting service has a key responsibility. This service not only needs to perform frame-cutting operation for each live video stream through the FFmpeg command, but also needs to save the generated image to the object storage OSS, and write the frame-cutting information into to Kafka. In this way, the downstream frame capture service can pull the frame capture information from Kafka, and obtain the address of the image in the OSS from the frame capture information, thereby completing the review of the image. In this architecture, Kafka is introduced to relieve the load of the audit service during peak business hours through an asynchronous processing mechanism.
​

​

Using function computing for the app, users do not need to purchase and manage infrastructure such as servers, but only need to write and upload code. Function Compute automatically prepares computing resources, runs tasks flexibly and reliably, and provides functions such as log query, performance monitoring, and alarming. With Function Compute FC, you can quickly build any type of applications and services and only pay for the resources that tasks actually consume. The architecture of the previous frame capture service can be migrated to the function computing platform with only minor adjustments to enjoy the value of serverless.

2. Function computing image real-time processing service

​

Take Sina Weibo as an example, the rapid growth of business has caused a great impact on Weibo's original technology system. How to not affect the user experience and increase the server cost investment is the top priority of Weibo developers in the case of the rapid growth of users, the surge of hot event traffic and the obvious peaks and valleys of requests.

The elastic expansion, event triggering, and pay-as-you-go features of Function Compute fit perfectly with Weibo. Sina Weibo decided to use function computing to deploy image processing services.
​

Sina Weibo deploys image processing services using the features of function computing's elastic expansion, event triggering, and pay-as-you-go, stores images uploaded by users in object storage, and writes functions to implement personalized image processing. When a Weibo user obtains a picture through the client, the request is sent back to Function Compute through Alibaba Cloud CDN. The function downloads the original picture from the object storage and processes it into the picture of the expected specification in real time according to the client type (such as different mobile phone models). , and return the result image, which greatly reduces the storage cost.

Function Compute's millisecond-level scaling of computing resources ensures that applications can still guarantee stable latency when hot events occur, and the user experience is completely unaffected by the number of visits. Weibo realizes continuous cost savings by running the image processing service through Function Compute. It is no longer necessary to reserve a large number of idle machine resources in advance to smoothly handle the traffic surge brought by business peaks. At the same time, because there is no need to maintain complex machine states, engineers can concentrate on working with product teams to increase business value instead of spending time on management infrastructure.

Summarize

Serverless technology will undoubtedly take on more responsibility, allowing users to build applications faster and better. Using the serverless architecture can cover many scenarios, here are just a few scenarios for image processing. Less is more Entrusting things to reliable platforms (such as cloud vendors), so that developers can focus more on their core business values, is the philosophy that Serverless has always advocated.
​

（END）

For more content, pay attention to the Serverless WeChat official account (ID: serverlessdevs), which brings together the most comprehensive content of serverless technology, regularly holds serverless events, live broadcasts, and user best practices.