In the IT infrastructure cloudization special session of the Alibaba Cloud Summit Developers Conference on July 24, Wang Guangfang, a senior technical expert of Alibaba Cloud edge computing, made a major upgrade of the edge node service, and discussed with the audience on the thinking and edge computing in the 5G era Technical practice.
In the 5G era, what kind of edge computing do we need?
With the completion of the issuance of the first batch of 5G licenses, a large number of terminal devices are connected to the network, data is transmitted through the network, and calculation and processing are performed. The service characteristics of low latency, large traffic, and large connection The challenge is very large. The traditional cloud and end architecture are not enough to support business development, and edge computing will play an important role.
The 5G standard defines ultra-low latency, large traffic, and large connections. The delay defined in the 5G standard is ultra-low latency. The air interface delay is less than 1 millisecond. Edge computing sinks the calculation to the closest to the device and the terminal. At the edge, such as mobile base stations and converged computer rooms, the original communication nodes are turned into computing nodes, which solves the delay consumption caused by multi-level network forwarding and achieves real low latency. At the same time, in order to cope with large bandwidth and high concurrency, the business logic and applications are also sinking based on the edge infrastructure, reducing the pressure on centralized processing.
On the whole, the edge computing required by an enterprise needs to provide three aspects of capabilities. The first is network and resource coverage, ensuring massive edge nodes and flexible infrastructure to cover terminal access, and the second is providing basic computing power at the edge. , Including computing, storage, network, security, scheduling and other basic capabilities. The third is to establish an edge remote operation and maintenance system to support rapid deployment, upgrade and monitoring of edge applications.
After the cloud and network infrastructure issues have been solved, another key task is to evolve the business architecture. Business needs to decide which center applications need to be sunk and which terminal capabilities should be moved up to achieve optimal performance and cost solutions? The typical cloud-side architecture scenario is the CDN. The CDN caches the content at the edge to complete the traffic convergence, achieve low latency for terminal access, and reduce the bandwidth of the center back to the source and optimize the cost. The evolution of this business architecture is relatively complicated. The mission of ENS is to make this matter simple, so as to accelerate the process of business architecture upgrade in the 5G era and support the incubation of more innovative applications. For example, in the cloud game scenario, if you put the business logic processing, data processing, and screen rendering of the game on the edge, the terminal does not need to have too many configuration and performance requirements, so the terminal cost can be greatly reduced, which is Breakthroughs and changes brought about by architecture evolution.
Alibaba Cloud Edge Fusion Computing Platform ENS
Edge Node Service (ENS) is defined as an edge fusion computing platform. Fusion refers to the integration of edge infrastructure such as MEC, IDC and other resources. There are also integrations in terms of calculation form, provision form, and operation and maintenance. The ENS can solve some of the problems faced by traditional self-built edge facilities:
First, the business elasticity is poor. Before using ENS, the self-built edge node cycle takes 1-2 months. The business elasticity is relatively poor, and the ability to respond to sudden demand is weak. Resources will also be idle after the peak of business. And after using ENS, it can realize the distribution of the marginal computing power at the minute level nationwide.
Secondly, in terms of cost, self-built nodes need to do resource procurement, hardware supply chain management, comprehensive operation and maintenance, etc., and the business startup cost is relatively high. The use of ENS requires no start-up capital and reduces the overall cost by 30%.
Third, it is reliability. We must face DDoS protection and host security issues on self-built edge nodes. At the same time, we must deal with the stability risks brought by software and hardware failures to the business. It takes a lot of energy to develop a complete solution. . While ENS provides high-reliability services, it can detect and handle problems in seconds.
Fourth, in terms of operation and maintenance, multi-node remote operation and maintenance is very challenging, and ENS can improve operation and maintenance efficiency by 30% through systematic and automated operation and maintenance.
ENS comprehensive upgrade-the first full-coverage edge computing infrastructure is released
. The convergence of mobile networks and fixed networks is basically completed on the provincial capital backbone nodes. In the operator's basic network, the provincial capital network nodes are hubs, so based on the provincial capital It is very important to have edge networks covered by nodes with high-quality network performance and stability. Relying on this, enterprises can select resources to build their own edge business and obtain better edge basic network service capabilities.
The current edge resource coverage and flexibility are very important, which is also the core capability that edge computing needs to provide to enterprises. This time, Alibaba Cloud ENS announced a major upgrade and released the first global edge node service in China. The total number of all edge nodes of ENS exceeds 300. The average coverage of each operator in each province is 3, and multiple available edges can be established in the province. layout. At the same time, ENS global edge nodes also support more than 20 multi-line edge nodes. Covering the three major operators of all provincial capitals in the country, popular areas support coverage in third-tier cities. The provincial capital node is a hub node in the entire network. ENS provides customers with high-quality provincial capital node resources. Supporting any device access delay at the network edge is less than 5 milliseconds, which can help customers quickly sink core business based on this edge facility. , Build a wide range of edge applications.
In an interactive live broadcast scenario, the biggest business requirement is low latency. Because the anchor and the audience are distributed throughout the network, after applying ENS, the anchor and the audience's terminal can be accessed nearby. The anchor video stream is accessed at the ENS node, and the transcoding is completed in the node. After the transcoding is completed, the CDN is used. For content distribution, the entire live broadcast link is completed at the edge without returning to the cloud center, greatly reducing the live broadcast delay and the success rate of connection in remote areas.
In the typical traffic localization scenario of local video information analysis, the video stream of the camera is accessed, and the edge node is directly sliced and stored. At the same time, AI computing power is used to identify key information and return the structured information. To the cloud center. Similarly, the entire traffic is converged locally, and there is no need to return the cloud center. ENS provides a large-capacity, low-cost object storage solution, while providing GPU virtualization and AI computing power to play a role in smart public security, smart stores, smart communities and other scenarios.
To provide developers with efficient and minimal edge computing
for operation and maintenance Compared with central operation and maintenance, edge operation and maintenance requires remote operation and maintenance based on the public network environment, and simultaneous operation and maintenance of multiple edge node clusters scattered in various places is very challenging. Based on the above characteristics, in order to help O & M engineers and developers to better cope with the challenges, Alibaba Cloud ENS and the container team firstly released the cloud-edge unified edge K8s and container solution-edge ACK in the industry. The solution is based on Kubernetes and implements cloud-edge-end integrated application distribution, supporting application distribution and lifecycle management under different system architectures and network conditions. The API enables developers to easily implement DevOps, unified resource management and scheduling, image distribution, self-service operation and maintenance, and cloud monitoring, which greatly simplifies the complexity of operation and maintenance.
In a customer's private network video service business scenario, because there are many and scattered edge nodes, the operation and maintenance is difficult, the resource expansion and management is complex, and the efficiency is low, I choose to use edge ACK products to implement one-click application operation and maintenance. K8s' orchestration capability manages and configures multiple applications in a single node, and uses K8s for image distribution and upgrade, dynamic scaling, and dynamic migration to complete the unification of the center and edge O & M control plane.
At the same time, in order to improve operation and maintenance efficiency, the edge node needs to respond in time when it encounters abnormal events or failures. ENS also docks with cloud monitoring products to synchronize the data logs and events collected in the edge node to the cloud monitoring, which is formed by cloud monitoring Standardized docking, notification, and alarm with user systems. At the same time, the user's system can control ENS instances and resources through OpenAPI to achieve a closed loop of the entire operation and maintenance system. Cloud monitoring supports notifications of node network events, node cutover events, instance status and migration, supports custom threshold alarms such as node network MTR indicators, instance resource occupancy indicators, and can be flexible through channels such as nailing, email, SMS, and telephone Reaching users, fully empowering ENS with minimal and efficient operation and maintenance.
