It's hard to imagine that the sudden New Coronavirus Pneumonia (COVID-19) completely disrupted the operation rhythm of domestic enterprises, but the unprecedented wave of online resumption of work unexpectedly ignited almost all the products that focus on overnight. However, a "short-term collapse" does not mean that the platform is defeated, and there are many technical issues behind cloud video conferencing that are worth exploring.
In April 2016, the domestic SaaS industry has not yet entered the second adjustment period. Wuzhao announced the first multi-person video conferencing function on Ding Nai. And Zoom is flourishing in the international market. Looking back at that time, many people should not realize that at that time, the third generation of SaaS was ready to change. Whether it's domestic nails or North America's Zoom, they are quietly capturing the battlefields of the previous generation of traditional video conferences such as Polycom and Webex.
At the same time, the surge in demand for cloud video conferences caused by New Coronavirus Pneumonia (COVID-19) has stimulated the development of the domestic cloud video conference market. The company's collaborative office platforms such as Dingtalk and WeLink launched free open services during the epidemic. The Tencent Conference provided remote collaborative services for teachers and students of Wuhan University, while Wuhan Union Medical College and other units enabled Good view to conduct remote consultations.
Some people even said that the entire cloud video conference industry has accelerated for 5 years. Compared with the previous period, the number of concurrent users, new users, and business traffic ... This almost explosive growth reflects the trend of the domestic cloud video conference market. However, this does not seem to mean that the service capabilities of enterprises will also increase, and this tests the speed of innovation and response of various enterprises.
Redefining audio and video
Although the cloud video conference system has many technical difficulties to solve, the core behind it is still real-time audio and video communication. In summary, there are three main technical factors.
The first is network transmission technology, that is, how to adapt to complex network environments to improve self-adaptation. This is actually based on the characteristics of the cloud video conference system. No dedicated line, any network, any device can be accessed, but the problem is that the network is unstable. How to improve the audio and video quality under the influence of network packet loss?
The second is the audio and video processing algorithm, which actually includes two levels, processing technologies such as noise reduction, and guarantees the smoothness and clarity of the video.
The third is multi-end convergence technology, whether it is web page, phone, APP access, or the opening of a wider range of application scenarios such as surveillance video systems, VR / AR equipment, drone video, etc., which are also of concern to many service providers.
However, from another perspective,
cloud video conferencing products also integrate all aspects of AI, cloud computing, security and other capabilities. Just as technologies such as beauty reduction, background blur, and anti-packet loss are also technical difficulties that are currently outstanding and difficult to break through in cloud video applications.
Due to the huge amount of information involved in video conferences, especially in educational and medical scenarios, sometimes sensitive and inappropriate content may be involved. The Alibaba Cloud video cloud team will also work with the Alibaba Cloud security team and the Alibaba Institute of Technology Collaboration, with regular banned text screening capabilities.
The normalization challenge of cloud services
The CPU is heavily occupied, many functions can only be used on mobile phones, there are experience issues such as chicken ribs, video stutter during high concurrency ... Domestic cloud video conferencing products have recently been spit by many users. In particular, the server downtime incident on the first day of the national resumption of work on February 3 and the online start day of February 17 has attracted much attention. The sudden outbreak of online office demand has led to a surge in access traffic. What kind of support is provided as a cloud video conference system in the rear area?
Judging from the network signal challenges of high peaks and unpredictable peaks of
the New Coronavirus Pneumonia (COVID-19) epidemic, Lou Jian, senior technical expert of Alibaba Video Cloud, summarized the following two points:
1. Rapid business expansion and short-term online launch of 10,000 machines to quickly deploy and service business applications are a big test for the operation and maintenance and service system.
2. Node scheduling and quality assurance capabilities under large business volumes: First, the increase in the number of nodes and equipment, and equipment problems that previously appeared with small probability, become high probability events in front of massive nodes; second, if the machine is due to hard disk and memory problems Hang up, need to monitor and intelligently cut at any time, and ensure that audio and video services are not perceived.
Facing such a challenge, Lou Jian and his team started from two aspects of audio and video:
The first is the "hearing" of voice: in terms of noise reduction and echo cancellation, a deep-tuned voice learning network is used to provide intelligent noise reduction for different conference scenarios such as open offices, open outdoors, and closed conference rooms. In addition, the stability and performance of the echo cancellation algorithm will be greatly affected due to the differences in the characteristics of the media collection characteristics of the actual mobile device. Through the establishment of a cloud feature library, it will continue to cover and collect new mobile devices in the market to continuously enrich compatibility.
The second is the "look and feel" of the video: self-developed real-time narrowband + SCC content coding can occupy a smaller bandwidth in the communication structure and effectively improve the fluency of the video; at the same time, SCC content coding, through the optimized coding of text and texture content, Make sure the text of your documents is clearer in video conference sharing.
It is worth noting that the outbreak has brought about a surge in user demand and allowed service providers to expand capacity. It can only be “relaxed” until the last two days. According to official data, Alibaba Cloud has continuously expanded 100,000 servers for Dingtalk.
But this does not mean that cloud services have only played a role in elastic expansion in supporting front-end collaborative office and video conference applications. The influx of users during peak periods for message processing, document delivery, and conference applications also severely tested the team's ability to resist pressure.
After a sudden increase in network traffic in a single area, we will schedule network traffic in areas with sufficient bandwidth and use a transmission algorithm to ensure that the experience does not decline. This will make full use of network resources across the country, and at the same time ensure that there are multiple, free-to-use Business reuse.
Since video conferencing is a daytime business, at night, these idle machines and bandwidth resources are used for other entertainment network services; at the same time, transcoding is idle, and during the low business hours in the morning, these computing resources are reused for large-scale media conversion. Code to reduce file storage and transfer size.
Non-negligible security issues
The New Coronavirus Pneumonia (COVID-19) epidemic has "fired" almost all software / systems with video conferencing capabilities. According to data published by many service providers, the number of users has skyrocketed, even once. Expand cloud server capacity and perform day and night maintenance 7 * 24 hours.
