Shulou(Shulou.com)06/02 Report--

This article shows you why to choose Kubernetes, the content is concise and easy to understand, can definitely brighten your eyes, through the detailed introduction of this article, I hope you can get something.

Choosing K8S is right, and Sextant can help you make good use of K8S.

Docker has swept the IT world like a whirlwind since 2013. Docker refers to container technology, which literally means container.

The small container is not simple. Someone once wrote "Container changes the World", saying that he has changed the world economy, and there are a long series of complex and interesting phenomena around it.

Containers are shipped on board, and the ship's helmsman sends them all over the world. The Docker of the IT world is installed in the Linux system environment of the data center, and it is up to Kubernetes, the helmsman, to decide the future of the container.

Some people say that Kubernetes comes from a good background (Google genes), and its work is good, but it is not easy to manage. As an old captain, several IT managers have come up with Sextant to tame Kubernetes,Sextant.

It literally means sextant, and it's also a matter of sailing at sea. It seems that the feelings of technicians are not distant stars, but the sea that can be reached.

Today we will mainly introduce Sextant (please ignore the meaning of the first few words of this name). This is a very serious thing made by several technicians from Baidu, percent and Cloud Voice, as mentioned above is to manage Docker containers, so what is Sextant?

The relationship between Sextant and Kubernetes (K8S) is like the relationship between RedHat, Suse, CentOS, Ubuntu and Linux. Sextant is a release of Kubernetes (K8S). The use of Kubernetes is as complicated as his name. These technical experts add many features on the basis of Kubernetes (K8S) to facilitate installation and deployment.

There are too many problems to be solved in using Kubernetes

Why do data center users use Docker,Kubernetes (K8S) and Sextant?

From a business point of view, when the business needs to go online, we need to apply for the required resources in time, release resources when the business is completed, and prepare for the next business when there is no business demand. In this regard, Yunzhisheng technology VP Li Xiaohan has deep feelings, he said, although our scale is not large, but there is also a great demand for flexible resources.

In terms of the organizational structure of the data center, there is the concept of business cluster in the data center. A cluster runs only one kind of application, and each cluster is relatively independent, which ensures the resources required by the application, but affects the utilization. It affects the consistency of data migration and flow within the data center.

The appearance of the container provides the possibility to solve these problems. The characteristic of the container is that the application is containerized and put inside the container, which solves the problem of environment dependence, so that it can be migrated easily.

But only containers are not good, and clustering management tools are needed. Only scattered containers need to be managed uniformly to have value. In technical implementation, we need a distributed operating system such as Kubernetes (K8S), so that all jobs are scheduled by the distributed operating system, and the distributed operating system can break the isolated business clusters in the data center and standardize the clusters.

The rapid development of big data's artificial intelligence has increased the data center's ability to process data and resources, and intensified people's demand for improving resource efficiency. Due to the same pursuit, this has brought big data's percentage point and Yunzhisheng's AI system together.

The common systems for managing containers are Mesos, Kubernetes (K8S), and YARN. Why choose Kubernetes?

Kubernetes originated from Google, many developers of Kubernetes, but also developers of Borg, the distributed operating system used internally by Google. The largest and most systematic application of distributed operating system is Google. As we all know, the overall utilization of Google's cluster CPU has been about 70% since a decade ago, while the overall utilization of CPU in many data centers is less than 20%. Google can do this. It must have a lot to do with the distributed operating system. As far as I know, the former Google employee, the current several people cloud CEO Pu return to start a business to do Docker is to see Google's advantages in this regard.

Liu Yiqian, vice president of technology at Kubernetes Community, said that Bug is often encountered on Kubernetes (K8S). Often, a Bug is detected in the morning, and when I want to go to the community to submit it in the afternoon, I find that this Bug has been resolved. Compared with other technology communities, the Kubernetes (K8S) community is more dynamic and popular.

Shortcomings of Mesos and YRAN

The cluster management in the data center has gone through many stages, from manual management to automatic management. Dr. Wang Yi, a senior scientist at Baidu Silicon Valley AI Lab, introduced the development of management at the container technology conference, and pointed out the problems and shortcomings of Mesos and YRAN (as shown in the figure above). In addition, he believes that Swarm is an effort for Docker to enter the cluster management, but it is too much affected by the stand-alone, and there is a slight deficiency in the cluster management.

As a result, Dr. Wang chose Kubernetes and successfully persuaded Liu Yicheng, who was doing technical selection, to research mesos and yarn at that time, and wanted to use yarn in January this year. After discussing with Wang Yi, he said that Dr. Wang recommended K8s. He said that at least at the abstract level, it is much higher than K8s, and it naturally combines well with Docker. Based on YARN, there will be a lot of unnecessary trouble.

So we saw that Dr. Wang Yi of Baidu, Liu Yigeng of percent, and Li Xiaohan of Yun Zhisheng all voted for Kubernetes.

Wang Yi concluded that Kubernetes changed the daily work flow of engineers, the development team delivered Docker image instead of just executable files, and Kubernetes simplified integration testing, eliminating the need for operation and maintenance work such as release, packaging, deployment, execution, monitoring and garbage cleaning.

Another kind of efficiency improvement brought by Kubernetes lies in reducing the difficulty and cost of developing parallel computing frameworks such as MapReduce, Spark, Storm, etc. Developers no longer have to apply several mature open source frameworks to write their own distributed programs, but can easily develop their own parallel computing frameworks based on Kubernetes.

Compared with the previous generation technologies such as Mesos and YARN, which were designed to schedule offline jobs, Kubernetes can schedule both offline and online jobs. Kubernetes provides a platform for cooperation, which is different from the former, and it is also the technical basis for this cooperation between big data's percentage and Cloud Voice's AI system.

On the Kubernetes platform, development teams can simultaneously run percentage points of EFK technology to collect and manage data, as well as Yunzhisheng's large-scale AI system to extract knowledge from data and improve service quality. Coupled with the common pursuit of both sides to improve the efficiency of data center operation and maintenance, there is Sextant.

According to reports, this project was launched on April 26 this year, and the joint development was led by Dr. Wang Yi of Baidu, and a virtual team was set up by percent and several technicians from Yunzhisheng to jointly develop it outside of daily work.

It took three or four months to develop, and the first version of Sextant was released about a month after it was launched.

Choosing K8S is right, and Sextant can help you make good use of K8S.

According to Liu Yicheng, vice president of technology, 1000 machines are concentrated in external applications. The utilization rate of CPU is 18% and the utilization rate of Hadoop is 100%. As big data has to collect a large amount of data, in the face of the rapid growth of data, he is thinking about making full use of resources, thinking about saving costs by starting with technology, and building the application foundation of web and Hadoop.

In a percentage of usage environments, data deletions need to be done frequently. Liu Yixiang envisions hadoop all a thousand machines. The advantage of this is that there are more resources to improve the utilization of CPU, and because it breaks the dedicated cluster and alleviates the problem caused by data migration between multiple clusters. After Sextant deployment, it will be combined with ceph to provide this capability.

Fang Liang, CTO of Pratt & Whitney Financial, said that the financial industry has begun to be decentralized due to the emergence of Internet finance. Due to the lack of credit measures in China, big data's information collection and risk control model have become particularly important. How to organize these computing resources so that they are flexible and scalable? The answer is based on the changes brought about by container technology. The Kubernetes distributed operating system has really developed because of its combination with big data and AI industries. Big data, artificial intelligence is combined with the latest technology.

Sextant is an early open source project. With the evolution of Kubernetes technology, Sextant still needs long-term maintenance development. These developers of Sextant welcome more people to join them. For many people who want to use containers, consider K8s and Sextant.

The above is why you chose Kubernetes. Have you learned any knowledge or skills? If you want to learn more skills or enrich your knowledge reserve, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.