The future of platform as code is kubernetes extension

translate ​ person |  Li Rui

Reviser |  Sun Shujuan

Infrastructure is code ： Where are from  

In recent years , With Docker The emergence of and the increasing popularity of containers , Infrastructure is code （IaC） The concept of . Infrastructure is code （IaC） It was originally used to connect virtual machines 、 Specific infrastructure such as network and storage API, Then gradually expand to include the operating system and Kubernetes And their configuration and reinforcement strategies . When viewing such as Terraform Such infrastructure is code （IaC） Tool time , They even support the deployment of workloads .  

What has not changed is people's initial understanding of “ Namely code ” Why are you excited . It all boils down to the familiar tools used in software development （ Editor 、CI/CD etc. ） And the process （ Code review 、 Version control, etc ）, And apply them to the lower layers , And make it descriptive 、 Repeatability 、 Shareability , And equally important automation .

Platform is code ： The way forward

The next step is to extend this concept and its advantages to the platform that you want to provide for developers . The goal is to build something like platform as a service （PaaS） The system of , Abstracts the infrastructure and enables engineers to focus on their code . In an ideal situation ,PaaS The system will get self-service without disturbing the developers 、 Standardize and share common best practices and the benefits of some type of security and implementation compliance .  

However , Typical PaaS The system has some common pitfalls that people should avoid .  

First ,PaaS The abstraction of often leads to artificial restrictions , And as software and developers grow and mature , They will encounter more restrictions . Now? , For traditional or closed PaaS System , This leads to exceptions and is modeled as a bad solution . secondly , Conventional PaaS Often accompanied by a high supplier lock-in rate . Third , People have an unpopular question to ask ： Is it really enough to use one platform ？ Whether the enterprise's data science engineers need to use the same platform as their e-commerce teams ？  

Kubernetes Offer help

“Kubernetes It's a platform for building a platform .” If people are familiar with or understand Kelsey Hightower or JoeBeda etc. Kubernetes Thought leaders , I may have heard of their statement .  

Again , Suggest Kubernetes Can actually become the preferred platform for more than just containers . in fact , This may be one of the things needed to finally achieve the desired goal of the people's envisioned platform, that is, the code .  

Kubernetes The benefits of （ For example, as a coordinator and a unified interface ） Forms the basis of this argument .Kubernetes As a coordinator, it brings effective coordination methods , Think of it as a more powerful declarative paradigm . It allows encoding of operational knowledge , This is more flexible and future oriented than building this knowledge into any form of script . Besides , State based storage is typical IaC Typical disadvantages of storage and state in tools .  

Kubernetes As a unified interface , It has brought a universal API, It has built-in authentication 、 Rate limiting and auditing . and , The API Has become the standard for cloud native workload management , And with its native extensibility , Yes KubernetesAPI Your familiarity translates into API Expand . because Kubernetes In recent years, it has achieved success and become popular , From continuous integration （CI）/ Continuous delivery （CD） The tradition of IaC To modern times GitOps Method tools are widely supported .  

Last , Many enterprises have extended... For many use cases API, Included in Kubernetes Clusters are defined internally 、 The first consensus was reached on the common abstraction of applications and infrastructure services .

Kubernetes The building blocks of the platform —— Beyond container orchestration  

First , There are clusters API project , It's in 1.0 Is ready for production . For beginners ,Cluster API Yes, consensus API Upstream efforts , Manage any infrastructure in a declarative way Kubernetes The life cycle of a cluster . If this sounds like a normal to the user API, don't worry , It includes the following API The work of , To generate clusters on many infrastructure providers , Including very large-scale servers and common on premises solutions .

After checking the cluster , Next up are the applications and workloads in the cluster . For a fully functional cloud native platform , A basic set of observability tools is required 、 Connection tool 、 Tools that make up the developer pipeline , You may even need some additional security tools or service grids . Now? , As a community , At least you can agree with Helm As a general packaging format . however , How to put these Helm The diagram is actually deployed to the cluster , Especially in a multi cluster environment （ As cluster management becomes easier , This is becoming more and more common ） It is still an area where consensus has not yet been reached . If the enterprise has joined GitOps The trend of , image FluxCD Such tools have some abstractions , for example HelmRelease, Can help .GiantSwarm Developed a project called app-operator The open source Kubernetes Expand , It expands Helm, Added multi cluster function and multi-level coverage for configuration , Thus, the pain of configuration management is reduced in the use case of deploying the application cluster to the cluster . It also allows for more diverse data to be included in the deployment process （ Such as test results and compatibility information ） Be prepared .

Another type of resource that people cannot ignore is the services of cloud computing providers . ad locum , See that most super large-scale developers are developing their own native Kubernetes Expand , In this way, they can generate what they call first-party resources . First party resources are Kubernetes Managed databases in , You can connect from cloud native workloads . Another very interesting method is Crossplane, It's an open source Kubernetes Expand , Enables users to assemble services from multiple vendors through the same extension , Provide an abstraction layer , Reduce lock-in to actual providers .  

These are just basic extensions ; Great progress has been made in this field , More and more projects or in the background Kubernetes, Or extend it publicly to their use cases . In the context of building the platform as code , In particular, some more specific frameworks and extensions need to be mentioned , These frameworks and extensions cover specialized but common use cases , for example Kubeflow Project MLOps/AI and KubeEdge Edge calculation of .  

Future vision and challenges  

It is still in Kubernetes Extensions and platforms are the early stages of code . Most standardization work is also at an early stage , But it is rapidly moving towards consensus and production readiness .  

The most important problem that needs to be solved now is the extended user experience of this kind . This is not limited to improvement API Validation and default values for , It also involves extended discovery and its document level . Besides , Once some of these standards are used closer to production , As a community, we need to pay attention to maintaining API The composability of , And promote interaction without tightly coupling the system . Last , There are many Kubernetes Debugging and traceability in extended complex systems can still be improved .  

However , One thing is for sure ,Kubernetes Will establish itself as the preferred interface between infrastructure and cloud native technologies . Besides , More standards will be established , More tools support and integrate with these Standards .  

All in all , The forecast for the future is ,Kubernetes Will become the standard cloud native management interface . It is the consensus of a unified community API. Of course , Users are still free to use the tools of their choice , But the unified open source interface ensures that users will not be locked .  

With the help of Docker And the container , People create a situation where the workload is transient . Using this technique , Not only can this concept be extended to Kubernetes colony , It can also be extended to the entire developer platform , Or many platforms provided to users .

Original title ：​​The Future of Platform-as-Code is Kubernetes Extensions​​​, author ：Puja Abbassi​