According to the data reported by relevant research institutions, the current market size of application containers will increase from $762 million in 2016 to $2.7 billion in 2020.
Obviously easy, the great flexibility of introducing container effectively promotes its adoption rate, which makes enterprises increasingly rely on this technology. At the same time, container technology has gradually become the strength substitute of virtual machine. Forrester has pointed out that 58% of developers plan to use containers in the next year or are planning to use containers.
In conclusion, using containers can help enterprises improve efficiency, reduce costs, and even have more reliable security guarantees. These easy to package and lightweight components can run together with other components in the same virtual machine.
In addition, the strong adoption of container also allows developers to work faster and better by creating virtual “sandbox”, so as to complete the writing, management and operation of software code, which can be completed without affecting the server or virtual machine (VM) running other applications and systems.
Based on this, CSDN cloud computing has specially planned a series of articles on container service inventory. From the professional and objective perspective of CSDN neutral technology community, it is intended to explore the current “container service” in the eyes of cloud service providers and provide some referential suggestions for developers to select appropriate container services, so as to help them realize the innovative application of container technology.
For this reason, we interviewed several cloud service providers that provide container services. In this issue, we specially invited Zou Hui, the product director of Tencent cloud container, to share in detail the insights on the development of containers and the aspects of tke, the star product enterprise level container cloud service platform.
Zou Hui, product director of Tencent cloud container
The concern and practice of new technology is not overnight
According to a Jing, Tencent cloud invested in container technology quite early, and tke is the star product of its container service, which is often summarized as “providing a high-performance container management platform with container as the core and highly scalable based on native kubernetes”.
Specifically, tke can seamlessly connect Tencent cloud’s capabilities in computing, network, storage, and security, and effectively expand kubernetes plug-ins such as network, GPU virtualization, and specific CRD resources. It not only helps complete functions in effective deployment, resource scheduling, service discovery and dynamic scaling, but also improves the convenience of business management under large-scale container cluster. In the future, Zou Hui, Tencent cloud container product director, also stressed that as the core part of Tencent cloud in container products, the deep optimization of tke function layer and the evolution of surrounding ecology will become a crucial thing for the team.
When it comes to tke, we will unconsciously think of the innovative practice of Tencent cloud’s native application arrangement platform. In addition to covering the container service level, the well-known development and operation platform, service mesh, no service suite, big data and artificial intelligence solutions based on the above products are all included. It is understood that at present, the platform has successfully carried hundreds of millions of containers in Tencent, supported the smooth operation and iterative upgrading of many massive businesses, and also successfully enabled the Internet, e-commerce, government affairs, energy and other fields and directions.
“In fact, Tencent has paid attention to the development of cloud native and container technologies for a long time, which is destined to be one of the most representative and promising technology directions in the future, so we have invested a lot in it,” Zou concluded
exactly! According to a Jing, Tencent cloud pays much attention to both products and community operations. From the product level, Tencent cloud has professional R & D teams in the fields of container, micro service and no server to support technology and business, and the launch of related products is endless.
“Compared with other cloud manufacturers, Tencent has researched container related technologies and services for a long time. Many successful businesses, such as games, wechat, and advertising, have chosen to run on container technology, which can be said to support billions of users,” he said.
From a technical point of view, Tencent’s internal container construction has also experienced a development path from unsmooth to mature. In this process, no matter in the function or stability of the container, or in the utilization rate of resources, there has been gradually in-depth discussion and research. In the future, it may be the key advantage of Tencent’s cloud container to choose to transfer these experienced technologies and experiences to the cloud and open them to more external enterprises or individual developers.
In addition to products and technologies, in the field of community, Tencent cloud chooses to strive to embrace the trend of open source technology in the past two years, open more and better internal architecture and experience, and even code to more users, and increase investment in cloud computing community at the same time.
“The development of containers in Tencent has gone through many stages, from the earliest use of CGroup for resource isolation and improvement of resource utilization to the later use of docker containers and a set of scheduling framework based on yarn. After the emergence of kubernetes technology, the container platform in Tencent gradually switched to kubernetes technology stack, and began to embrace cloud native technology in an all-round way. With the evolution of container technology, Tencent’s internal awareness of open source is also changing, and the investment in the open source community is very large. On the one hand, we actively participate in the open source community, on the other hand, we will gradually promote some good practices and technologies within Tencent to the community, bringing value to more people and enterprises. ” He added.
According to the current market observation, the technology of pro test container has successfully entered the development stage of mature stage. Among them, kubernetes is the representative, the version update is faster and faster, the function is more and more perfect, and the ecology is more and more perfect. In addition, we see a lot of monitoring, microservices and higher-level deployment, as well as a lot of AI and big data frameworks that have made a lot of deep-level functional evolution around k8s. Now the whole container technology has been widely accepted. Many enterprise level users are preparing and practicing many container technologies. The major cloud manufacturers also regard the container technology as an important opportunity point for the next cloud market technology explosion. In the community aspect, the cloud community represented by k8s is gradually growing into one of the most popular technology communities in the industry.
Referring to the technical points that Tencent cloud pays attention to in this context, Zou Hui said that the severless technology, which has attracted much attention at present, should be able to further the development of containers and effectively help enterprises obtain better and stronger capabilities on the way of micro services. In addition, the emergence of lightweight virtualization technology can make the performance of containers more stable and secure, and really release the value of containers in the multi tenant hybrid deployment environment.
We have to deal with all kinds of container fields
Although container technology is booming, there are still many problems in the process of enterprise landing. It is worth noting that at present, many container technologies in the industry are provided to users in an open-source way, but it is one thing to use the open-source technology in practical operation, and another thing to use it well.
“Especially for some small and medium-sized enterprises, there is not only a certain threshold for k8s technology itself, but also a greater difficulty in building a very good application on k8s, which can not only enjoy the convenience brought by k8s, but also ensure the reliability of the application. There are many points to be considered, which need more technical support and practice accumulation. In the process of using containers for many customers, due to the unreasonable setting of pod resources, the core and k8s related components are triggered after oom, and the node notready is finally triggered. What’s more, if the cluster resources are not reserved enough, an avalanche effect may occur. So how to discover, evade and recover these problems in advance is a key consideration for users in the process of using k8s.
Also for large and medium-sized users, it is worth exploring how to pull the more traditional complex architecture in the enterprise to the container service represented by k8s. It is mainly because in this process, in order to achieve containerization migration, large enterprises will inevitably expand new functions on the basis of k8s in order to adapt to the original habits or functions, such as disk isolation, network isolation, container IP remaining unchanged, etc., and need to migrate the traditional complex framework to the container.
So on this basis, Zou Hui proposed that Tencent cloud in the field of containers will add more abundant basic product forms. In addition to the standard k8s services currently provided to the outside world, the lightweight virtualization + virtual kubelet technology will also be used to provide the serverless k8s cluster such as k8s native interface to the outside world. In this elastic cluster, users only need to pay attention to the container resources when using k8s, so as to minimize the attention to the underlying resources. At the same time, Tencent cloud container service will also export some good internal practices of Tencent, such as TAPP application management (used to make kubernetes service and pod management more refined and release process more controllable), GPU virtualization and other technologies.
In addition, of course, there is also an important aspect to expand the use scenarios of containers, such as: tke combines with istio to provide grid services for users; tke products combine with big data and AI industry to create solutions with more AI or big data capabilities based on k8s + virtualization; in the game field, tke will also cooperate with Tencent’s game technology operation Department To share the typical experience and technology used in the game architecture containerization of Tencent and launch a containerized game solution.
In addition, in the container private cloud and hybrid cloud, Tencent cloud also launched the TKE enterprise container platform, which provides enterprises with private container solutions. Tke Enterprise Edition adopts the same architecture and management mode as Tencent’s public cloud container service, so when enterprises privatize the management of container service, they can also easily get through the container service on the cloud and get the same management experience to realize the hybrid cloud deployment; in the future, we will further expand the functions here to realize the hybrid cloud management.
Tencent cloud enterprise container cloud platform tke
In addition to the interview, Zou Hui also gave his own suggestions for the competition among container manufacturers. He pointed out that as container manufacturers are facing the market environment with more and more obvious demand and broad prospects, if they want to grasp the key tickets, the most important thing is to focus on the stability grinding of the bottom layer of the product, which is a very key point.
In addition, it is more necessary to complete the overall shape based on the container. For this, a Jing understands as follows: in addition to providing standard k8s services for use, it is more important to provide complete upper level services to solve the elastic and other series of needs, so that the whole becomes more valuable.
Finally, in the field of private cloud containers, it is necessary to continuously improve the delivery efficiency of containers and reduce costs to achieve a more standardized and streamlined user experience. A Jing thinks that this is not only a tip for container service manufacturers, but also one of Tencent cloud’s goals in the field of container technology.
AI, severless and service grid, fire technology needs to keep freshness
Referring to the performance of AI technology related containers, Zou Hui also detailed the practice in Tencent cloud. To be sure, at present, the new learning frameworks represented by AI are all looking to k8s one by one, because the deployment and scheduling capabilities at the bottom of the application container can facilitate the operation of the upper framework. In this regard, Tencent cloud has indeed accumulated a lot of experience.
Of course, there are also difficulties. For example, some teams may use GPU in the process of reasoning tasks or in the virtual process in the way they used to. They may buy some virtual machines, but compared with the actual operation efficiency of GPU card, it is very low, running and idling alternate with each other, and the cost is large; in addition, the performance of different GPU cards is different, and the underlying environment is heterogeneous It’s also a challenge for the AI team. Tencent has done a lot of technical optimization in these blocks (cost, performance, heterogeneous), such as GPU virtualization technology, GPU topology identification Wait, this is one of them.
“In addition, when some of our AI teams deployed the AI framework and services before, they relied more on human resources to deploy and manage the underlying resources; when combined with container services, these greatly liberated their operation and maintenance deployment work and brought a lot of convenience.” He said.
During the interview, Zou Hui was quite sure what role severless technology could play in the innovation and development of container technology? The concept of no service computing is put forward to further expand the product form of container service, so that users can really not pay attention to the underlying resources, such as when to expand the capacity of nodes, whether the nodes fail, how much the utilization of node resources, etc., so that users can free themselves from some tedious work and pay more attention to the business and architecture.
In addition, what about Tencent cloud’s technological innovation in service grid? According to a Jing, with regard to the service grid, it is also gradually promoted in Tencent. At the same time, it also encounters two biggest problems.
The first is performance. Because the current technology will do a set of extra hijacking, unpacking, analysis and repackaging processes for all capacities, Tencent cloud has actually done a lot of research on performance optimization, such as istio In addition, Tencent also has a lot of experience in using istio extension protocols. How to adapt these custom protocols to istio has also done a lot of work.
Before ajin’s release, Tencent cloud has released four cloud native technology products on the cloud native technology conference kubecon + cloudnativecon + open source Summit (original LC3), including the enterprise container service platform tke, container service grid tke mesh, serverless 2.0, and one-stop Devops platform. As a leader in the field of cloud computing, a Jing feels that many practices of Tencent cloud in the fields of containers, microservices and serverless in the future will really open the cloud era.
(reprinted from InfoQ)