Author | Alibaba Cloud Native
Source | Alibaba Cloud Native Official Account
Nowadays, cloud native has become a well-known buzzword, and it seems to be outdated without mentioning cloud native. Joining CNCF has also become a technological advantage that cloud vendors are proud of.
We have also seen various cloud-native definitions, such as "microservices + containers + continuous delivery + DevOps" from CNCF, as well as from different cloud vendors. In September 2020, Alibaba Cloud established the Cloud Native Technology Committee. Today I will try to talk about what is true cloud native from the beginning of cloud computing.
Cloud native in a narrow sense
Let us first review the origin of cloud native.
After the rise of cloud computing, the wave of containers represented by Docker has swept the industry and is widely favored by DevOps and standardized delivery. In this context, Google creatively put forward the concept of cloud native, and released the open source orchestration tool Kubernetes, which cut in from the perspective of unified deployment and standardization, and subsequently established the CNCF Foundation, which successfully made a breakthrough.
In December 2020, CNCF Ambassador Ian Coldwater said on social media: "Kubernetes has now deprecated Docker support. You need to pay attention to this and plan accordingly. This will ruin your cluster."
This shows that the cloud native strategy has made historic progress. From this perspective, equating cloud native with containers and service grids is an early understanding of cloud native in a narrow sense.
Cloud native
So, what is cloud native in a broad sense? The software, hardware, and architecture born from the cloud are truly cloud-native.
We believe that cloud native should be more from the perspective of customer applications. Applications deployed on the cloud must use one or more of the three types of capabilities that only large-scale public cloud practices can provide, namely elasticity and API automation. Features such as deployment and operation and maintenance; service-oriented cloud-native products, such as RDS, EMR, etc.; software-hardware integrated architecture born from the cloud. This is cloud native.
First of all, cloud-native applications are relative to the offline traditional enterprise IT physical machine environment. The cloud must be used to deploy and manage applications, make full use of the cloud's supply efficiency, and require global deployment, out-of-the-box, pay-as-you-go.
The ability of customers to obtain the cloud nearby is the so-called elasticity. We published a book last year, called "Elastic Computing Ubiquitous Computing Power", which talks about the underlying principle. The cloud is the only way to have this extreme elasticity. Native. Kubernetes provided on the cloud realizes cloud nativeness through API standardization for automated deployment and operation and maintenance.
Secondly, today we have seen the concepts of cloud-native database, cloud-native big data, cloud-native container, cloud-native middleware, cloud-native security, etc. These are all service-oriented cloud-native products that can be obtained at will on the cloud, and are traditional Services that are not available offline can help to obtain performance and cost enhancements. This is cloud native .
Particularly worth mentioning is the integrated virtualization of software and hardware represented by the Shenlong architecture. Virtualization has experienced pure software virtualization, general hardware virtualization, and now software and hardware integration virtualization.
Through the separation of computing and storage, the use of dedicated Shenlong chips to accelerate virtualization, so as to achieve high performance beyond physical servers. As the bottom layer of the cloud computing architecture, all middleware and application layers can obtain performance benefits from it without any code and architecture modifications.
In addition, the dragon architecture supports the resource capabilities of the same set of resource architecture and pooling, including ECI (Serverless Container), VM, bare metal, etc., so that all resources can be scheduled in a unified manner, thereby achieving the deepest flexibility. This is also the first time that the industry has realized a unified architecture and unified resource scheduling, and is regarded as a major technical contribution of Chinese cloud vendors.
In this sense, the Shenlong architecture is born out of the cloud and is a typical cloud native. Simply put, if there is no large-scale deployment scenario for cloud computing, the advantages of Shenlong cannot be fully utilized.
Therefore, the original intention of cloud native is to be native on the cloud. If it lacks the basic characteristics of cloud, it must not be cloud native. In extreme terms, if you buy two servers yourself, put them in the office, deploy containers on them, and then use Kubernetes to manage them, it is not cloud native, because of the lack of flexibility, API automation deployment and operation and maintenance capabilities, even the cloud It's not .
CNCF is more about advocating what components a system should use for architecture, so as to realize the Internet architecture and operation and maintenance capabilities, and achieve rapid iteration, flexible expansion, and efficient operation and maintenance. These capabilities require that these frameworks be cloud-based, and the combination of the two is cloud-native.
Many people will ask, is a certain open source software cloud native? This depends on its architecture. Take an open source database as an example. If its architecture is still aimed at traditional stand-alone and small-scale, it is not cloud native; if open source interface compatibility is maintained, cloud native transformation and redesign have been made, and operation, maintenance and control are fully utilized The basic capabilities of the cloud are cloud-native software and services, such as Alibaba Cloud’s PolarDB database.
Multi-cloud and hybrid cloud accelerate cloud native
The development of history is always tortuous, and because I believe it, I see it. When driving on the winding road, often only a limited field of vision can be seen .
In order to achieve higher cost performance, fault tolerance and efficiency, the industry has produced multi-cloud and hybrid clouds. Multi-cloud is when customers deploy business loads to different public cloud vendors, while hybrid cloud is more where customers deploy business loads to public clouds and private clouds.
Multi-cloud is a product of competition, although Amazon and Alibaba Cloud have always believed that one cloud is enough for users. The hybrid cloud comes from a different perspective. It is a cloud architecture built to satisfy customers' willingness to go to the cloud step by step, while meeting the regulatory compliance needs of some industries.
Whether it is multi-cloud or hybrid cloud, in order to reduce the difficulty of distributed business deployment and make it more acceptable to customers, cloud vendors will urge customers to unitize their business, because Kubernetes solves the problem of standardized deployment and migration.
At the same time, in addition to Kubernetes and containers, it often involves the use of public cloud services such as elasticity and acceleration, as well as cloud-native transformation of databases, big data, audio and video systems. It can be seen that Kubernetes accelerates the narrowly defined cloud native capabilities.
Cloud Native Chip
From the perspective of computing chips, cloud computing brings new application scenarios, which puts new demands on CPUs, that is, cloud native chips.
For example, Alibaba Cloud's Shenlong chip is a cloud native chip. It is the core of the entire device, because it allows the Shenlong server to have the characteristics of a virtual machine, including the interface of the virtual machine.
The Shenlong MOC card based on the Shenlong chip plays an important role in the entire architecture. It has a wealth of management interfaces and external data interfaces, which can help us realize these functions of the flexible bare metal server.
The diversification of CPUs is gradually becoming a reality . The first is the rise of Intel’s rival AMD. Although they are all x86 systems, their chips are cost-effective, and their performance and features have also advanced by leaps and bounds. Secondly, the development outside the x86 architecture is very rapid. With the exception of Ampere’s ARM server chips, The researched ARM system is becoming the choice of many cloud vendors.
AWS’s Graviton 2 self-developed chip is customized using a 64-bit ARM Neoverse core. The design idea is multi-core plus a single Thread load per core. Each core is designed to deliver the actual workload through as many, completely independent cores as possible. Strive to be the most energy-efficient processor.
Microsoft is also designing an ARM processor for data centers. This is a major strategic move by Microsoft. Microsoft believes that its own chips are more suitable for certain needs and have more cost and performance advantages.
Cloud-native CPUs are showing some common characteristics: the main frequency does not need to be particularly high, the power consumption is particularly low, independent running cores, the multi-core architecture can better support the application of the service grid, and the mature software ecology.
Use cloud native thinking to make good use of cloud native architecture
For cloud computing users, understanding what cloud native is is particularly important. Instead of simply using containers to define your own system, but in the design, development, and deployment cycles, using the original cloud native system will bring huge competitive advantages and fully enjoy the dividends of cloud computing.
Cloud native is not only a practice, but also a concept, bringing about changes in the times. The development environment of developers must also be born and grown on the cloud. Use remote tools like Wuying to make product design documents, development and debugging environments, deployment tools, testing processes, POC verification, resource consumption, etc., all based on the cloud.
Therefore, what cloud native brings is a change in thinking, a change in culture, and a new era of productivity. It far surpasses the Kubernetes standard interface defined by CNCF and is the standard way of using cloud in the future.
What is true cloud native
We are very fortunate that we are experiencing a historic journey into the cloud era.
Years later, a new generation of new humans will be even more fortunate. All computing environments they come into contact with are cloud-native. They are born to believe that cloud computing resources are ubiquitous, inexhaustible, and inexhaustible. They don't care about where and how many cloud resources are. Just like we use tap water today, no one will bother to think about where the water comes from .
The software, hardware, and architecture born from the cloud are truly cloud-native; the technology born from the cloud is the cloud-native technology.