Kubernetes' learning path. Is there any "easy mode" Q recommendation for container hybrid cloud

author ｜ Zhang Yawen

In recent years , The hybrid cloud model, which combines the advantages of public cloud and private cloud, has gradually become the mainstream .Flexera released 2021 The cloud status report shows ,92% Our company is in IT Choose a multi cloud strategy in terms of Architecture , among 82% Enterprises choose hybrid cloud . With the application of hybrid cloud more and more widely , It is not easy for more and more users to complete the choreography in the cloud . although Kubernetes Has become the de facto standard for container orchestration and scheduling , however Kubernetes The operation is complicated , And only focus on single cluster tenant management , In multi cluster management , In particular, the multi cluster management involving cross cloud is not perfect . Besides ,Kubernetes Designing for cloud data center also has some limitations in edge computing scenarios .

face Kubernetes Existing problems and strong market demand for hybrid cloud . Major public cloud manufacturers have launched their own hybrid cloud Container Services , At that time , All kinds of products and solutions are dazzling . How to choose among the hybrid cloud container products with various advantages ？ Just figure out the technical route behind the hybrid cloud container service , It will help guide developers on the painful path of technology selection .

1 be based on Kubernetes Is the open source project fragrant or not ？

The current hybrid cloud container services can be roughly divided into two categories , One is based on Kubernetes Of , The other is not based on Kubernetes Self research project . because Kubernetes Has become the de facto standard for container orchestration and scheduling , therefore , The former is a well deserved mainstream . In recent years , Major public cloud manufacturers have also successively opened their own source based on Kubernetes Hybrid cloud container service . The technical routes of such services are mainly divided into two categories .

The first is in Kubernetes Do subtraction on the basis of , Create lightweight container choreography Services , More typical products K3s.

K3s By Rancher Lab On 2019 Lightweight launched in Kubernetes Distribution version .K3s The name of comes from K8s-5. In order to reduce Kubernetes Memory required ,Rancher K3s Deleted the old 、 Non essential code 、 Consolidate the packaging process that is running 、 Use containerd Instead of Docker As a container engine at runtime , In addition, introduce SQLite As an optional data store . So as to make it better fit CI、ARM、 Edge environment 、 Internet of things and testing these five scenarios .K3s All components run on the edge , So it doesn't involve collaboration , But from the perspective of production K3s Lack of cluster management scheme , Responsible for cross cluster application management 、 Monitoring etc. , But unfortunately Rancher These capabilities are not open source .

With K3s The lightweight container choreography products represented by are suitable for running in edge computing scenarios . The existing Kubernetes Distributions are usually memory intensive , It is too complex in the edge computing environment . These products reduce memory , So that it can be better deployed in edge scenarios , Besides , In the edge computing scenario , Enterprises need operation and maintenance management Kubernetes The number of clusters is very large , And usually only a small number of nodes , Therefore, operation and maintenance personnel need to be responsible for large-scale infrastructure . Orchestrate products through such lightweight containers , It can simplify the installation and operation steps of users to the greatest extent , Reduce the difficulty of operation and maintenance .

The second is to the existing Kubernetes Make some improvements to the architecture . Under this technical route, the more representative open source products are KubeEdge and OpenYurt.

At present Kubernetes Architecture , The whole control plane and data plane have high requirements for network , Therefore, this kind of technical route has made some improvements to the scene , Both products provide an agent layer in the middle to maintain and ensure a relatively stable application environment in an unstable environment .

KubeEdge yes 2018 An open source edge computing platform in , At present by CNCF Incubation projects . The project is in progress. Kubernetes Native container layout and scheduling capabilities , The extension realizes cloud side collaboration 、 Calculate the settlement 、 Massive edge device management 、 Edge autonomy and so on . From the perspective of architecture, the cloud （K8s master) Added Cloud Hub Components and types controller, And on the edge （K8s worker) Is to rewrite the native component EdgeCore Components , As can be seen from the architecture diagram EdgeCore Is based on Kubelet Reconstructed , To ensure lightweight , Cut the original Kubelet Part of the ability , Added a lot of ability to adapt to edge scenes . Although compared with Kubernetes,KubeEdge Indeed, it can better adapt to the edge scene , But due to differences in Architecture , There will inevitably be some impact , For example, the compatibility of cloud native ecology will decline .

OpenYurt Compared with KubeEdge Open source is a little later , But it uses a non intrusive approach to Kubernetes enhanced , That is to say Kubernetes Add on . In the cloud (K8s Master) To add YurtController ManagerYurtApp Manager as well as Tunnel Server Components . And at the edge (K8sWorker) I added YurtHub and TunnelAgent Components . Compared with Kubernetes,OpenYurt Improved edge unitization 、 Marginal autonomy 、 Cloud edge collaboration and other capabilities . At the same time, there is no loss of compatibility with the original ecology of the cloud . However, due to the OpenYurt Always add to the original architecture , Therefore, lightweight is difficult to guarantee , Besides , Native Kubernetes The problem that too many edge nodes may lead to system instability still exists , Not resolved .

be based on Kubernetes Is the open source project fragrant or not ？ Undeniable? , In specific scenes, such as edge computing , They made up for the original Kubernetes Deficiency . But while making optimization , New problems often arise . Because these two technical routes are in Kubernetes Make some improvements in the architecture , Therefore, it will have the impact of architecture differences to varying degrees . Besides , Because some projects are right Kubernetes Intrusive modification of the system , Follow Kubernetes The evolution of the community will face great challenges .

2 Keep the original flavor Kubernetes framework , Optimize performance

Since the improvement of architecture will bring the impact of architecture differences , So can we use the original Kubernetes framework ？Amazon EKS Anywhere by Kubernetes Ecology provides a new idea .

Amazon cloud technology was launched this year 9 It is fully launched in June Amazon EKS Anywhere.Amazon EKS Anywhere Completely native Kubernetes framework , Make no changes , Just add some management and maintenance tools on the original basis , Make it fully compatible and more convenient to deploy in the user's own data center .

Amazon EKS Anywhere yes Amazon EKS Latest deployment options , In the past Amazon EKS Infrastructure compatibility can only be achieved in Amazon's own cloud environment , But through Amazon EKS Anywhere Users can run on their own infrastructure Amazon EKS. therefore ,Amazon EKS Anywhere More suitable for those who already have 、 Business users who intend to continue to use large-scale local infrastructure .

Amazon EKS Anywhere Help simplify local configuration through default component configuration Kubernetes Creation, operation and maintenance of clusters , At the same time, it provides tools that can realize the automation of cluster management . It is based on Amazon EKS Distro Built on the advantages of , The latter is for Amazon cloud technology Amazon EKS The same... That provides support Kubernetes Release version . Amazon cloud technology supports all Amazon EKS Anywhere Components , Including integrated third-party software , This allows users to reduce support costs , There is no need to maintain redundant open source and third-party tools .

Besides ,Amazon EKS Anywhere Provide users with Amazon EKS Consistent local Kubernetes Operating tools . Developers can take advantage of EKS The console passes EKS The connector looks at all the components running anywhere Kubernetes colony （ Include EKS Anywhere colony ）.

although Amazon EKS Anywhere It has many advantages , But there are some limitations . According to the current official statement of Amazon cloud technology ,Amazon EKS Anywhere Also applicable only for local use VMware vSphere Deploy on your own infrastructure , For other deployment targets, such as bare metal environment , Will be in 2022 Provide support in . in addition Amazon EKS Anywhere The positioning of is used for the deployment of local data center , There is no support for other public cloud deployments .

3 Kubernetes Hybrid cloud container services outside the ecosystem

however , be based on Kubernetes R & D Amazon EKS Anywhere Although a lot of work has been done on the use threshold , It has the advantage of low intrusion at the architecture level , But fundamentally ,EKS Still based on Kubernetes Of , The characteristics are very distinctive . For developers , There is still a high entry threshold .

If the enterprise team is right Kubernetes Not very familiar with , Or no time to research 、 Study Kubernetes, How to deal with the container arrangement and governance in the mixed cloud environment ？ Amazon cloud technology was launched this year 5 Month launched Amazon ECS Anywhere, Maybe it can be another answer .

According to introducing ,Amazon ECS Is the first container orchestration service managed by Amazon cloud technology . Provide users with a set of easy-to-use control plane , You can use virtual machine instances （Amazon EC2) Or no server at all （Amazon Fargate) Form to easily run various container workloads , At the same time, it realizes native integration with other hosting services of Amazon cloud technology , And then provide service grid , logging 、 Enhancements such as indicator capture .

Amazon ECS Anywhere Appearance of function , Enables users to deploy all kinds of services in non Amazon environments Amazon ECS Mission . On this basis , Customers can use the same set of easy-to-use management to define and manage all resources in the cluster within a specific Amazon region , Regardless of where the cluster is located , How the execution environment .

Compared with Kubernetes The most important difference is Amazon ECS More powerful , And easier to use . Even though Kubernetes The ecosystem of open source products is very prosperous , But the problems are very complex , The learning curve is steep ,Amazon ECS It is very simple and easy to use, and it is closely combined with many products on the cloud , It's very convenient to use .

Besides ,Amazon ECS Anywhere, It is very suitable for use in scenarios where edge computing or user computing resources are relatively limited , Very light 、 flexible , Not much for hardware , Or resources 、 Particularly strict requirements on the network , So there are many application scenarios .

from 2021 From the year onwards , All developers in the cloud , The general trend of cloud edge end integration can no longer be ignored , There will be more than 50% The data run in all types of IoT On the terminal ,Amazon ECS Anywhere It is the fitting and best fit of this trend .

Besides , Developers just need to Amazon ECS Anywhere Deployed in container services outside Amazon's infrastructure , Use External Start type , Can make the same Amazon In the region Amazon ECS The cloud services in the cluster can interact with each other in the same environment , also , The external container load can also give full play to the local way in the physical location to serve the system , This enables large-scale processing of internal video files and other high-intensity work that was difficult to complete in the cloud in the past .

because Amazon ECS Anywhere Emphasize the complete neutrality of infrastructure , So as long as the operating system developed by the developer can run Amazon ECS and System Manager agent , You can register all kinds of machines or equipment , The extremely low registration threshold helps developers easily outside the Amazon environment , Including the deployment of micro service use cases at the edge , Completely free from the operational requirements of the data center .

4 Will open source be the mainstream of container in the future ？

So looking back , How to choose a hybrid cloud container service ？ At the level of container management , If the team is right Kubernetes Experienced , And has a mature architecture system , Of course, you can choose to face... At the cluster level Kubernetes Deep customization .

But if we need to promote R & D more efficiently , And with the cloud 、 Combined with the general trend of cloud edge end Integration , Such as Amazon EKS Anywhere、Amazon ECS Anywhere This kind of products are undoubtedly better at getting started , Easier to maintain , And shielding the underlying complexity .

It was also mentioned that , The future of containers is absolutely open source , Customization based on open source software , It's for the future .

indeed , For the ecology around the container CNCF The panorama of cloud native technology is really huge , Explain the prosperity of the whole open source field , And many technologies are really driven by open source , It is the result of the contribution and development of everyone in the open source field ; But on the other hand , After all, the container layer belongs to the middle layer , It can optimize our application development for the upper layer , Promote its iteration speed . For the lower layer, it is inseparable from the support of the basic resources of the lower network computing layer , At this level, it can't get rid of , So at this level , Some technological innovations made on the public cloud , Better able to run applications on containers .

For the future of container technology , The author believes that it will develop in three directions .

First , The field of containers has become a de facto standard technology , Include Docker as well as Kubernetes Will continue to develop rapidly , More traditional enterprises will migrate in ;

secondly , In the past, containers were deployed in a relatively centralized way , For example, in the past, it was either deployed on the public cloud , Or use it in your own data center , With the architecture of hybrid cloud , There will be a big change ;

Third ,Serverless No combination or integration of server and container , It is also the development trend in the future . In terms of the so-called cloud generation maturity , This is a more mature way , Be able to move into the field of no server , Therefore, container based serverless technology is also an important direction of development in the future .