Introduction: In the 10th issue of Cloud Computing Intelligence Bureau, Yun Yue, an expert of Alibaba Cloud Products, fully revealed the product design background, product functions, application scenarios and value of the new product — View Computing online, leading netizens to explore the new “View” world.
Recently, the 10th issue of cloud computing intelligence agency, Ali cloud product expert Yun Yue on the new product – view computing product design background, product functions, application scenarios and value of a comprehensive online reveal, leading users to explore the new “view” world.
In 5G era, view data has become the most important carrier of information digitization
In Yun Yue’s view, the large-scale digitization of information can be roughly divided into three stages: the first stage is the digitization of text, the second stage is the digitization of speech, and the third stage is the large-scale digitization of video and picture data.
There is a very interesting picture in 1994 of Bill Gates sitting on 330,000 sheets of paper with a CD and proudly announcing that a CD could hold the contents of the entire paper. The capacity of this CD is in the MB level. In 2010, Mr. Jobs held an iPhone that claimed to be able to hold the contents of tens of thousands of compact discs, roughly in gigabytes.
(Image from the Internet)
When 5G arrives, what characteristics do we see in the entire view data? To sum up, the first is video and picture data, which are widely generated on various terminal devices such as mobile phones, on-board terminals, unmanned aerial vehicles, game consoles, etc. The second feature is that the data level will reach ZB/day; The third characteristic is dispersion, the terminal device in any location may produce video and picture data; The fourth feature is that for the whole data, the value density of video and image data is relatively low. All of these characteristics will bring a very big challenge to the whole data digitization. How do we meet this challenge? How to use cloud computing to more easily meet the arrival of the above scenarios? View computing products are designed to undertake such business scenarios.
View computing – Cloud PaaS services for service (view) terminal computing
The positioning of Ali Cloud View Computing product is to provide the nearest connection, storage and computing PaaS service for the view terminal. This product combines the characteristics of Ali Cloud Edge Computing node and public cloud, and carries out a cloud-oriented design for the massive view terminal devices. Among them, edge computing nodes provide cloud computing nodes with city-level coverage and can be closer to device terminals. View computing also combines the characteristics of public clouds to make big data analysis, data persistence storage, and application deployment easier.
In order to achieve the design effect of the product, Ali Cloud has designed the view computing product architecture in three layers:
The first layer is view-oriented data processing and constructs the architecture of edge computing nodes.
The entire node including base layer of physical resources, hardware selection, basic network architecture have made specific design, such as to view for large-scale data analysis, the ability of providing data cache to the nearest, the architecture of data cache storage (cycle) is the ability to do special design optimization, provide a higher ratio of performance data to the ability of memory. Through periodic storage, the view data can be stored in the edge computing node for 1 day, 10 days or a month. When these data are to be used for further data-oriented application, the data can also be synchronized to the public cloud region for further data persistence. In the intermediate process, the high-density value data generated by AI reasoning calculation in the edge calculation section can also be returned to the central region through the edge network for further big data processing.
At the same time, the node architecture also comes with some basic computing power, including transcoding screenshots, etc.
The architecture design of the second layer of layered computing based on endside cloud collaboration
The so-called end-to-side cloud collaborative layered computing is to effectively connect devices, edge computing nodes and public clouds through the view computing PaaS platform, so that data can be calculated at appropriate locations.
To achieve this effect, view computing has the ability to connect to the device’s terminals in order to control the terminals. When some data needs to be processed at the terminal, the data can be initially processed at the terminal. At the same time, the nearest computing node can connect the data generated by the device terminal nearby for calculation. The generated data can also be calculated through the public cloud for further large-scale application and analysis.
In this way, the layered computing of the whole end-to-side cloud collaboration is formed, which can ensure the efficiency of the whole computing.
The third layer is the multi-node cooperative nearby computing.
Because view data is generated at the terminal, customers prefer to have cloud nodes close to the terminal for data processing, so view computing needs to have the ability to cooperate with multi-node computing, so as to ensure that all terminal devices are connected to a edge computing node, and processed and analyzed at the edge computing node.
View computing builds the capabilities of location awareness and platform scheduling. The capabilities of the platform can be location-aware, allowing developers to focus on the entire business process without having to focus on the location of specific physical nodes.
Based on the above three-layer product architecture design, view computing has three product capabilities:
First, the ability to connect the device terminal.
Terminal connection and data uplink management can be carried out through the one-click cloud open protocol platform that comes with view computing. At the same time, you can also connect the data of the device terminal to the cloud by standard means, such as GB, RTMP or ONVIF. To accommodate developer flexibility, View Computing also has some custom design capabilities in the Connectivity section, which means developers can let Data Computing connect device terminal data to the cloud under a custom protocol.
Second, nearby data storage capacity.
View computing provides the ability to cache nearby views so that periodic data can be stored nearby. It also provides the ability to return data to the center for data persistence and large-scale data analysis.
Third, AI computing power
View computing integrates the AI computing capabilities of Damo Academy, and also comes with some basic computing capabilities, such as screenshots, transcoding, etc. In order to make the whole video data processing more flexible and convenient, view computing also provides a custom interface, developers can develop their own or third-party algorithms integrated into the view computing platform, to achieve more flexible business development. At the same time, view computing also provides open, unified interfaces and open protocols to help developers achieve more flexible integration.
In order to achieve the above product capabilities, the underlying support includes scheduling, management, application hosting, monitoring, operation and maintenance and other technical modules, in order to achieve more efficient collaborative management, higher stability and reliability.
Ultimately, view computing can host a multi-billion tier of terminal view data cloud, allowing customers to do simple and efficient development out of the box, based on an open interface.
Typical application scenarios for view computing
For all kinds of device terminals, customers of view-based computing products only need to do business process, data application development and final application deployment. Other view data connectivity, storage, and computing power can be carried over from the view computing platform.
When it comes to specific scenarios, Ali Cloud has implemented landing practices with partners in the following five scenarios:
The first is the road computing scene, which is based on the landing practice of the cloud on the data of the highway view. The final effect is that the entire vehicle traffic becomes safer and more monitorable. Using view computing to build an intelligent high-speed system, the traffic video can be processed in the nearby cloud, the video delay can be reduced to 10 milliseconds, the response time of traffic incident video AI computing can be increased by 70%, and the efficiency of road safety incident processing can be improved by 80%. This truly achieves the traffic situational awareness in the whole domain, and the service level has been greatly improved.
The second is the education scenario, in which the typical cloud classroom transforms the offline classroom into a digital classroom. Through view computing products, remote online lectures and online learning can be realized. At the same time, view computing can also do further AI analysis and application of view data to help digital classroom and view data of digital classroom, realize AI computing ability, improve students’ class efficiency, help teachers better monitor the education effect, and gradually improve their teaching effect through feedback;
The third is the new retail scene. Through view data, the purchase link and supply chain link of the retail scene are connected to help the retail scene improve the efficiency of sales and the efficiency of supply chain. Of course, there is a typical scene in the supply chain is logistics, view computing can help logistics scene, from the equipment, from the delivery of goods to do a full link of the whole process of monitoring, improve management efficiency, at the same time let consumers enjoy a more convenient shopping environment, help the new retail industry digital upgrade;
The fourth is public safety, including public safety, food safety, bright kitchen and bright stove video scenes; Overlay cloud and AI capabilities in video-based localization to improve management efficiency and reduce storage costs;
The fifth is the home entertainment scene, which can be further improved through the enabling of view data, such as cloud games and AR/VR.
View calculated product value
First, view computing is based on edge computing nodes, which has the ability of covering and processing nearby. The whole network has lower cost, higher access flexibility, and low latency.
Second, the data processing of view computation is hierarchical. First, the data can be calculated roughly on the terminal; Secondly, the data can be actuarial calculated on the edge computing node, and the value data of higher density can be extracted by further structuring. Thirdly, these data can also be transmitted back to the public cloud for large-scale data application and analysis.
At the same time, the data storage of view computation also supports layering. On a device terminal, data can be temporarily stored; When the business needs to do periodic storage, the edge computing node can be selected for further periodic storage of data. However, when part of the data is more important and needs persistent data storage, the data can be stored in the public cloud for further persistent storage through the scheduling capability of the data computing platform.
The benefits are lower costs, as well as an increase in flexibility and ease of expansion.
Third, another feature of view computing is the location-free PaaS layer cloud service. All development interfaces are uniformly exposed to customers and developers through the View Computing product, and behind the product are the massive devices connected to the view terminal, the nodes connected to the edge computing, and the public cloud. Therefore, the data processing location can be achieved without sense, so that the appropriate node to do the most appropriate and efficient computing services.
The benefits of this kind of product design are lower network cost and higher development efficiency.
Fourth, view computing not only provides an open protocol of one-click cloud to facilitate convenient connection and access to the entire device terminal, but also opens a programmable interface to help developers make their own definition of access protocol. As for the overall AI computing capability, it can not only integrate the AI capabilities of Aliyun Damo Institute, but also achieve more subdivided scene AI computing through a custom way. The result is more flexible business operations, more efficient development and lower costs.
future
At the end of the online sharing, Yun Yue expressed his hope that more ecological partners (node level and algorithm level) could develop and implement together with Ali Cloud, so that more business scenarios could use various convenience brought by view computing. At the same time, it is also expected that view computing has more intelligent scheduling and more intelligent management capabilities to help massive device terminals realize very convenient nearby access, storage and calculation, so as to mine a broader value of view data. This article is the original content of Aliyun, shall not be reproduced without permission.