The author what tube wave source | | alibaba cloud native public number

With the development of 5G, IoT, live broadcasting, CDN and other industries and services, more and more computing power and services begin to sink closer to data sources or end users in order to obtain good response time and cost. This is a computing method that is significantly different from the traditional central mode — edge computing.

However, while the scale and complexity of edge computing are increasing day by day, the shortage of operation and maintenance means and operation and maintenance capacity finally began to be overwhelmed. In this context, “cloud, edge, and end integrated O&M collaboration” has become an architectural consensus. The process of cloud side convergence is also being dramatically accelerated through cloud native enablement. Under such a trend, it is urgent to introduce the concept of cloud native and transform the operation and maintenance management mode of edge application comprehensively.

This article is compiled from the author of Ali Cloud container service technology expert, OpenYurt author and one of the start-ups He Linbo (Xinshen) on January 26 in Ali Cloud developer community “Tuesday Open Source Day” live sharing, will stand in the actual scene perspective, explore the fusion of cloud native technology and edge computing challenges. The architecture and industry practice of cloud native edge computing platform based on OpenYurt are introduced in detail.

Click back to see the full video: developer.aliyun.com/live/246066

What is Edge Computing?

Edge computing refers to the deployment of workloads on the edge of a computing approach compared to traditional centralized general-purpose computing. In recent years, edge computing has been very hot, mainly because 5G, IoT and other services and scenarios are developing faster and faster, including more and more intelligent terminal devices, resulting in more and more demands for edge computing services. If all the processing is in the center, it will be difficult to meet the growth of large-scale edge smart devices. Edge computing is now widely used in various industries, such as automobiles, transportation and energy. In summary, edge computing is about getting the computation closer to the user or to the data source.

1. Edge computing Top-level architecture

Industry – defined layered structure of edge computing, citing Gartner, IDC.

The hierarchy defined by Gartner is shown below: Endpoint > Near Edge > Far Edge > Cloud > Enterprise.

  • Near Edge: a non-standard server or device, located closest to the end.

  • Far Edge: standard IDC, which can be divided into three types: IDC (main), MEC, CDN, etc. Relatively speaking, the computing capacity is relatively strong, such as the operator’s machine room, cloud service provider’s cascading machine room and so on.

  • Cloud: Public Cloud or proprietary meta service, characterized by centralized and centralized resource management.

The hierarchical structure defined by IDC is shown in the following figure:

  • Heavy Edge: Data center dimensions; Centralized computing platform; CDN, self-built IDC.

  • Light Edge: low-power computing platform, suitable for industrial control, data processing, transmission and other Iot scenarios.

As can be seen from the figure above, Gartner definition and IDC definition are actually interdependent and interrelated. In addition, edge computing and cloud computing are not substitutes, but complementary and interrelated.

2. Edge computing industry trends

The trend of edge computing industry can be explained from the following three aspects (dimensions) : the first is the business of the industry, the second is the structure of the industry, and the third is the scale and change of the industry.

Trend 1: the convergence of AI, IoT and edge computing

In recent years, the combination of edge computing with AI and IoT is very frequent. With the increase of the number of edge intelligent devices, all data or videos are sent back to the cloud for processing. The overall cost and efficiency are very inappropriate, so the demand for AI processing or IoT processing directly near the device is increasing. The AI, for example, will do training on the cloud or in the central cloud, and then do reasoning on the edge, and there’s a lot of that. According to the survey:

  • By 2024, 50 percent of computer vision and speech recognition models will be operating at the edge.

  • By 2023, nearly 20% of servers used to handle AI workloads will be deployed on the edge.

  • By 2023, 70% of China’s iot projects will include AI capabilities, pursuing real-time, bandwidth reduction and data compliance.

  • By 2023, 75 percent of Enterprises in China will process iot data at the edge of the network.

Trend 2: Cloud extension, IT decentralization, facility autonomy, edge hosting

Edge computing and cloud computing are mutually complementary and interdependent. Taking it a step further, edge computing is actually an extension of cloud computing to the edge, extending some of the capabilities of the cloud to the edge. First, IT business is required to be decentralized on the edge. In addition, because edge business or facilities are autonomous, they have certain control ability and edge hosting ability when the network between cloud and edge is disconnected. The future architecture trend will evolve towards the development path of cloud extension, IT centralization, facility autonomy and edge hosting:

  • Hybrid cloud – By 2023, 10% of enterprise workloads will run on local data centers and edge resources.

  • Decentralization – By 2023, more than 30% of new infrastructure will be deployed in marginal locations.

  • Facility autonomy – 50% of core enterprise data centers and 75% of major fringe IT sites will change operations by 2024.

  • Edge Hosting – By 2022, 50% of companies will rely on hosting services to improve performance and ROI based on edge AI.

Trend 3:5G and edge computing explode new growth

In recent years, the rapid development of 5G is a new growth tipping point for edge computing. With the number of edge apps expected to grow 800% by 2024, you can only imagine the growth that will follow. Typical application scenarios will include Internet of vehicles (autonomous driving/vehicle-road collaboration), smart power grid (equipment inspection/precise load control), industrial production control, intelligent medical treatment (remote B-ultrasound/remote consultation), etc.

3. Current situation of edge computing

The edge cloud is driving a rapid rise in management complexity

With the growth of the form, scale and complexity of edge computing, the operation and maintenance methods and operation and maintenance capabilities of edge computing are increasingly difficult to meet the innovation speed of edge business. “In the future, enterprises are fully pursuing super scale, super speed and super connection”, which further aggravates the complexity of operation and maintenance management. The edge cloud contributes to a rapid increase in management complexity in four main ways:

  • First, the sharp increase in the number of intelligent Internet terminal equipment; The demand of data and business sinking increases.

  • Second, the scale and business complexity of edge computing have increased, and new businesses such as edge intelligence, edge real-time computing and edge analysis have emerged. The centralized storage and computing mode of traditional cloud computing centers can no longer meet the requirements of edge devices for timeliness, capacity and computing power.

  • Third, it is difficult to coordinate the cloud side and end, and there is a lack of unified delivery, operation and maintenance, and control standards. Moreover, it is difficult to control the security risks of edge services and edge data.

  • Fourthly, it is difficult to support heterogeneous resources, and challenges to support different hardware architectures, hardware specifications and communication protocols, as well as to provide unified standard services based on heterogeneous resources, networks and scales.

Cloud edge integrated edge cloud native

1. What is cloud native?

Definition of cloud native: Cloud native is an open, standard technology system. Based on the cloud native technology system, users can build and run a set of business systems with high flexibility, good fault tolerance and easy management. The whole technology system has many hot technologies, such as Cloud Native, Serverless, Kubernetes, Container, Docker, etc., which are widely used in the industry.

Now major cloud manufacturers and cloud service providers are investing heavily in cloud native, and cloud native is becoming more and more the entrance for users to use cloud computing capabilities. The cloud native technology system can help enterprises maximize the capabilities of the cloud and maximize the value of the cloud.

2. A rich family of Yunyuan products

Take Aliyun as an example, the product family of Aliyun is mainly divided into three parts, as shown in the figure below:

  • The first is the new application payloads, including data & Intelligence, distributed applications, and DevOps, which are now hosted natively through the cloud.

  • The second block consists of Serverless, container Orchestration, a new business system.

  • The third block includes: public cloud, private cloud, edge cloud is a new resource hosting system.

3. Cloud side integrated cloud native infrastructure

Cloud edge integrated cloud native infrastructure is a cloud native system that does control and edge autonomy on the cloud. As shown below:

On the side of the center, it can provide the management and control ability and production ability of the original cloud center. For example, Kubernetes+ storage /+AI/+ big data and other capabilities can be provided in the center. These capabilities of the center sink to Edge computing through control channels, such as standardized CDN, Infrastructure, Edge, ENS, or the equipment gateway of smart factory, smart park, building, airport and so on on the right side of the figure. On the edge, you can access all kinds of devices, such as sensors, videos, controllers, etc., and support all kinds of communication devices. This creates a cloud native infrastructure that integrates the cloud side with the cloud side.

Cloud computing is good at processing and analyzing non-real-time data with a relatively long cycle that requires massive and scalable storage capacity, while edge computing is derived from cloud computing. It is good at real-time processing and analyzing local short-cycle data. The relationship between cloud computing and edge computing is not substitution, but synergy. Close combination of the two can better meet the matching of various demand scenarios.

4. Cloud edge integrated value

The concept of cloud native was first proposed in 2013. After several years of development, especially since Google led the establishment of CNCF in 2015, cloud native technology began to enter the public’s attention and gradually evolved into DevOps, continuous delivery, microservices, containers, infrastructure, Serverless, FaaS, a collection of techniques, practices and methodologies. Cloud native accelerates the integration of multi-cloud and cloud-side. The value of the integration of cloud-side is:

  • First, it can provide users with the same functions and experience on any infrastructure as on the cloud, so as to realize the integration of cloud side and end applications.

  • Second, the isolation of the container can be used to ensure the security of the services running on the edge by using the system’s ability of flow control and network policy.

  • Third, through containerization, the support of heterogeneous resources can be well adapted through decoupling between containers and resources.

5. Difficulties in the fusion of cloud native and edge computing

With the growth of the form, scale and complexity of edge computing, the operation and maintenance methods and operation and maintenance capabilities of edge computing are increasingly difficult to meet the innovation speed of edge business. In the future, enterprises are striving for “super scale, super speed and super connection”, which further aggravates the complexity of operation and maintenance management.

What problems should be solved by cloud native and edge computing convergence? In the process of actual problem solving, the following four key points are summarized:

Firstly, the scale and business of edge computing are complex, and edge resources are scattered in different regions. Life cycle management, upgrade, expansion and shrinkage of edge applications within each region and internal flow closed-loop are all facing challenges.

For example, in the CDN scenario, there may be hundreds of computer rooms across the country, and the resources or machines in each computer room may be different, and the traffic of the services running on the machines may also be different. At this time, it is very insufficient to use native Kubernetes’ workload to manage, which will form a very big challenge, easy to make mistakes, and the whole operation and maintenance efficiency is very low.

Second point: The cloud side network connection is unreliable. Generally, the cloud and edge are connected through the public network. Under the influence of some objective factors, the network between the cloud and edge may be disconnected, which poses a great challenge to the continuous operation of edge services. Because in the case of network disconnection, the node will leave the cloud control, and the Pod will be expelled under the native K8s. However, in actual cases, edge services must continue to run whether services or machines are restarted. So the edge needs some autonomy.

Third point: It is difficult to coordinate the operation and maintenance of the cloud side. Because the machines on the edge are deployed inside the user firewall, the public network cannot be actively connected. Therefore, some K8s native operation and maintenance capabilities that need to pull data from the center cannot be used, and manufacturers lack unified standards for delivery, operation and maintenance and control, and it is difficult to control security risks of edge services and edge data.

Fourth, it is difficult to support heterogeneous resources. It supports different hardware architectures, hardware specifications and communication protocols, and challenges to provide unified standards based on heterogeneous resources, networks and scales.

OpenYurt Edge computing cloud native platform

CNCF Edge cloud project OpenYurt: an intelligent open platform extending native Kubernetes to edge computing.

1. Edge autonomy and center (cloud) control

The OpenYurt architecture is a very simple cloud-side integration architecture. As shown in the figure above, there are two parts on the cloud: blue parts are some components of native K8s, and then orange parts are components of OpenYurt. Every node on the Edge, every node on the Edge Note also has a blue part and an orange part, and the blue part is also a native K8s component, or some cloud native component of the Settings, and the orange part is OpenYurt component.

As you can see, OpenYurt is zero modification and non-invasive to K8s or to the cloud native architecture. OpenYurt project is the industry’s first non-invasive enhancement of K8s, an edge computing cloud native platform. Other edge computing cloud native projects, more or less, may be modified or tailored to K8s, which is the biggest difference of OpenYurt, thus ensuring the standardization of OpenYurt.

  • OpenYurt can keep pace with Kubernetes version upgrades.

  • The concept of non-intrusive, OpenYurt and mainstream cloud native technologies, such as ServeiceMesh, Serverless, etc., can be evolved synchronously.

OpenYurt entered the CNCF Sandbox in September 2020. It is a very neutral project, first in terms of technology and architecture, and second in terms of operation of the project.

The quality and stability of OpenYurt are also guaranteed. Within Ali Group, it is widely used and has managed millions of cores.

2. How does OpenYurt solve the fusion difficulties of native and edge computing

  • First, edge uniformization. In large-scale business, because the edge units are scattered, the business within the unit can be managed unitary and the flow can be managed closed-loop through edge unitary.

  • Second, the ability of marginal autonomy. In order to cope with the unreliability of the cloud side network, by adding an autonomous capability to the edge, the continuous operation of services can be guaranteed even when the cloud side network is disconnected.

  • Third, seamless conversion. The main purpose is to reduce the threshold of OpenYurt use, by providing a seamless conversion ability, so that K8s and OpenYurt cluster can be switched between a key, a command can convert a standard K8s cluster into OpenYurt cluster, reverse switching can also, this is the industry’s first ability.

  • Fourth, it solves the problem of cloud access to actively access the edge, providing the ability of cloud side collaboration to solve the difficult problems of operation and maintenance.

Each of these capabilities is described in detail below.

1) Unitization ability

It provides the ability to apply models in edge scenarios, mainly including the following three points:

  • NodePool manages nodes in batches in unitary mode.

  • Traffic management Supports traffic topology management of native services.

  • UnitedDeployment provides unitary deployment of the native APPs model.

Unitized mainly is the ability to provide edge scenario application model, the resources the nodes in the pool, can to each region of a node, a pool of management, on the edge of the above unit 1, if be a room in Beijing, the node can in Beijing pool inside, can these nodes are a mass of labels, and other functions of management, In this way, it is very convenient to manage the overall characteristics of the same batch of machines. UnitedDeployment This resource is based on node pools and manages node pool services by node pools.

Unit according to the above, for example, the deployment of two instances, the deployment of unit 2 three instances, the configuration is submitted to OpenYurt cluster, can automatically deployment information is distributed to the edge, then you can start the corresponding instance number, it has solved the unit management problem of each unit to operate independently, and from the Angle of a unified, UnitedDeployment can manage individual cells.

2) Marginal autonomy

Escort the continuous operation of marginal business, including the following two points:

  • YurtHub caches cloud data. When the cloud is disconnected, all system components obtain data from YurtHub.

  • Yurt-controller-manager Eliminates the problem of edge service expulsion when the cloud side network connection is unstable.

The edge autonomy capability escorts the continuous operation of edge services and ensures the continuous operation of edge services even when the cloud side network is disconnected. It mainly involves two components, one is YurtHub and the other is Yurt-Controller-Manager.

YurtHub is a component deployed on edge nodes in the form of containerization on each node. From the figure above, understand the processing process, request native components such as Kubelet, KubeProxy and Flannel, which are directly connected to cloud APIServer before. Now adjust to connect to YurtHub before forwarding requests to APIServer.

This advantage is when the request to come over, cloud edge network without break, have a health check module, will detect cloud edge network connectivity, if cloud edge network is normal, the request directly to the load balancing module, then select a cloud server forwarding in the past, the results back, one can return to a requester side, The other result data is cached on the local disk and persisted on the local disk.

If the cloud side network is disconnected and the node needs to restart, it can extract the local delayed data through the Local proxy and return it to a requiter, so as to restore edge services and ensure the continuous operation of edge services.

3) Seamless conversion ability

The seamless conversion capability is accomplished with the YurTCTLConvert component. It is mainly used for one-click conversion between standard K8s and OpenYurt cluster; Currently, the cluster supports the deployment of tools such as minikube, Kubeadm, and ACK.

In the case of conversion, because there are many nodes in the cluster, each node needs to be converted to the edge node, some yurthub static pod components, kubelet parameter modification, etc. As shown below:

Through the broadcastJob of OpenKruise, another cloud native open source project of Ali, a job like POD can be guaranteed to run on each node to complete the node-to-node transformation. At present, our Yurtctl tool has carried out a relatively complete verification on the cluster deployed by minikube, Kubeadm, ACK and other tools. We will support more types of cluster in the future, and welcome more interested students to contribute to the community.

4) Cloud-side synergy

As shown below:

The YurttunnelServer component is deployed in the cloud, and each edge node will deploy a Yurttunnel Agent. When Yurttunnel Agent is started, the ANP Proxy Agent module inside it Establish an encryption channel for bidirectional authentication between the cloud network and the ANP Proxy Server module. This channel is done by gRPC protocol.

After the channel is established, when the cloud visits the node, the IPtable Manager in Yurttunnel Server will import the request traffic of node access to Yurttunnel Server. The Request Interceptor Interceptor module intercepts the request, converts it to the gRPC protocol format, and forwards the request to the TunnelAgent on the edge, which forwards the request to Kubelet or POD. In this case, the cloud side operation and maintenance coordination ability can be got through. So that the native Kubernetes operation and maintenance operation ability, can run in OpenYurt cluster or cloud scenarios without perception. In addition, the cloud side operation and maintenance channel is based on gRPC protocol. By compressing Tunnel bandwidth, the cost can be greatly reduced and the traffic can be reduced by up to 40% compared with the original TCP communication.

OpenYurt case introduction

Case 1: Edge AI

The first case is the edge AI scenario, which is hema’s new offline retail business.

Based on ali cloud container service ACK@Edge product as the base, Hema Has carried out the transformation and upgrading of cloud native, and constructed a Digital full-link tianyan AI system of cloud, edge and end integration of “people and goods yard”. First, there is a cloud on the cloud of control surfaces, and then in the area of edge near stores, bought ENS node services, so you don’t have to oneself to store building room, and then through the cloud and integrated system or modeling system, the entrance guard control system deployment to the edge of the ENS services, with the inside of the store after monitoring video stream delivery, Then the business load is analyzed and the results are obtained. On the side of the control business system, the calculated results are returned to the cloud for analysis.

Based on public cloud and cloud + edge, the business architecture of Cloud Tianyan system, Ali Cloud edge convergence node ENS and Hema store physical field is realized at a low cost. It has strong flexibility, mixed resource management ability and cloud original operation and maintenance ability. Achieve 70% increase in service opening efficiency of new stores and over 50% resource cost savings; Sharing computing power. As shown below:

Case 2: Cloud on video

Cloud cases on video are widely used across the country, as shown in the figure:

Look from down to up, such as on the motorway, lightweight gateway or standard gateway, there will be some video shooting flow, the video shoot, to the near edge of ENS or CDN server, video monitoring, such as some provincial, city and county of room inside, do video management, gathering forward after processing, Upload the final results to the cloud control platform on the central cloud. Then, the cloud control platform can do a lot of processing, for example, we can cooperate with Autonavi to release some events or inform information, so as to form an integrated service management platform of cloud side and end.

Through the service management platform of cloud side and end integration, including application deployment/configuration, device/application status, and cloud realization of structured data, the overall operation and maintenance efficiency and control efficiency are greatly improved.

This is the share about OpenYurt. If you are interested in OpenYurt, please scan the code to join our community exchange group and visit the official website of OpenYurt and GitHub project address:

  • OpenYurt website:

openyurt.io

  • GitHub Project address:

Github.com/openyurtio/…

“Kubernetes and Cloud Native Application Open Source Practice Lecture”

4 cloud native and Kubernetes technology frontier topics live, 70 classic courses, 3 cloud native e-books, to “Kubernetes and cloud native application open source practice lecture”, and Ali cloud container technology experts together, will be popular container open source projects and cutting-edge cloud native application landing practice! Click direct “Kubernetes and cloud native application open source practice lecture”!