Shulou(Shulou.com)05/31 Report--

What does the PAAS model OAM refer to in the original cloud era? many novices are not very clear about this. In order to help you solve this problem, the following editor will explain it in detail. People with this need can come and learn. I hope you can get something.

With the rise of kubernetes, many companies have the ability to build Paas platform, but the application of Paas platform construction is basically different forms, a hundred flowers blossom, and OAM in the author's view is the application of Paas platform construction of kubernetes, the future factual standard, today let's talk about OAM.

1. 1001st Paas platform

Before we talk about OAM, let's talk about the traditional operation and maintenance development, the evolution from the operation and maintenance system to the operation and maintenance platform, and the problems that may be encountered.

1.1 initial stage

In traditional operation and maintenance development, basic components such as CMDB, automation, monitoring, publishing, logging and process systems are basically standard. Metadata is stored through CMDB, atomic operations are provided automatically, and then continuous delivery is achieved through release. This stage can usually be called stage 1.0.This stage can provide certain support capabilities, and the main goal of operation and maintenance in this stage is to meet internal requirements. Export continuous delivery capabilities (just delivery, most companies'CI processes are controlled by testing, praise the team's own experience)

1.2 platform stage

The main goal of the platform phase is to achieve self-service operation of R & D as much as possible through the integration of operation and maintenance systems. Typically, it is based on ITSM to achieve the linkage of the above process platform, R & D to fill in fixed work orders, and then through process scheduling, integrate the current operation and maintenance subsystem, realize the self-service operation of a scene, reduce the participation of operation and maintenance, and provide R & D efficiency. It may get through to some other teams in the company, such as database, testing, middleware teams, let's call it stage 2.0.

1.3 Service phase

Service-oriented Paas is mainly through the service-oriented capabilities of the underlying infrastructure and operation and maintenance systems, and there is a highly unified goal within the company, which begins to change to the cloud transition phase. Each team provides highly cohesive service-oriented capabilities, while externally providing the platform lifecycle management capabilities.

Here we want to understand the difference between service-oriented capabilities and system functions, for example, the release system provides dozens of parameters, allowing users to provide random definitions. I understand that this may only be a function, because the higher the degree of freedom of the user, the less uniform the process specification of the platform, and if you let a user use the system, you have to know dozens of functional parameters repeatedly. God, good luck! As for the service-oriented ability, what I understand should provide users with release strategies, as few control parameters as possible, standardized processes, and specific complex choreography capabilities are completely integrated into the platform and have no perception of users. Just call it stage 3.0.

1.4 1001 options

At this stage, the builders of the platform usually consider what the next big direction of the platform is. From the current operation and maintenance product category, we can be divided into three directions: efficiency devops direction, Paas direction, operation and maintenance portal, but we all have a very consistent goal is to improve R & D efficiency and achieve application lifecycle management.

1.4.1 Operation and maintenance Portal

The goal of the operation and maintenance portal is mainly to cover the three stages of testing-> release-> operation and maintenance. By integrating the capabilities of the operation and maintenance side, such as cmdb, monitoring, publishing, logging and other systems to achieve unified operation of the application, it is usually combined with the company's CMDB to achieve unified management. Its advantage is that it meets the operational requirements of operation and maintenance, and individuals understand that it belongs to a stage of construction relative to devops and cloud. The main emphasis is on integration.

1.4.2 devops performance

The typical representative of the performance devops direction is Ali's cloud effect, from requirements-> Development-> Test-> release-> Operation and maintenance-> Operation to achieve full link coverage. In most workers, it is very good to get through testing-> release-> Operation and maintenance-> operate this link, but usually the company will not do this. As long as you have touched it slightly, you will know how difficult it is to get through the stages of requirements-> development-- > testing.

1.4.3 Paas

Paas direction in addition to testing-> release-> operation and maintenance-> operation of this link coverage, it is very important to provide the basic capabilities of the cloud, of which the key capabilities: flexibility, multi-tenancy, self-service, pay-on-demand, of course, there are many prerequisites, for example, you have to be flexible so that the company has resources, pay-on-demand also want to do billing, but if you want to build a system, you must think about it. What if we have to have a hybrid cloud in the future?

1.4.4 Operation

With regard to the understanding of operation, on the operation and maintenance side, I currently understand that it is to maintain stability, reduce cost and improve efficiency. Stability should be fundamental to operation and maintenance, and the main objectives of operation should be the latter two.

Cost reduction means that the platform can determine and reduce the operating cost through some mechanisms. A typical manifestation is whether the business uses resources reasonably. No matter in K8s or traditional vm, there is a problem that R & D has applied for 4h8g machines. Is it really reasonable? If we can establish a set of operational standards, for example, we can measure the resource utilization of the business according to its resource usage in the past week, month, quarter, or even year, is it possible to reduce some costs?

Efficiency may be a difficult indicator to measure, because after being unable to build a platform, it is more likely to say how much you have improved than before, more likely from how long it takes users to use the platform, such as launching applications. How long did it take from application creation, resource application, and online delivery? For example, if the company provides unified scaffolding, scaffolding related standardization construction, to open CI, CD, monitoring, logging, security and other functions, then R & D only need to pull down the project from git, and then code development, and finally go online, apply for resources?

Of course, this is only the author's idea, there is no practice, interested can talk about ideas together, after all, this may be easier than AIops these may be easier to land.

1.5 Perplexity of choice

In fact, whether in performance or Paas, we are all doing the same thing, that is, application lifecycle management, but we will find that the definition of application in different directions and different companies is absolutely different, and there is no unified specification for application lifecycle hosting, and the operation and maintenance R & D teams of most companies are not large. How to design a high cohesion, low coupling, scalable and distributed application Paas platform, the hairline estimation is much higher.

In the current cloud native era, you can quickly build a company's container cloud platform based on the native capabilities of K8s Paas (Saas) cloud. We may build our own App and then combine Service, DEployment and other resources to describe an application, and then adapt to the log / monitoring. In fact, everyone is imperceptibly following a common standard, which is actually provided to you by K8s, but for application dependence. How do you describe such information as mysql and redis? How to describe this kind of middleware? This is today's protagonist OAM.

2. First acquaintance of OAM

The full name of OAM is Open Application Model, and in terms of its name, it defines a model and implements an OAM-based model that I think aims to define standards for cloud native applications.

Open: supports heterogeneous platforms, container runtimes, scheduling systems, cloud providers, hardware configurations, etc., regardless of the underlying layer

Applications (Application): cloud native applications

Model (Model): define standards to make them independent of the underlying platform

The description of this paragraph is quoted from Mr. Song's article, refer to Appendix 1

Here I would like to talk about my understanding. We all have a painful point of platform design. A shutdown that has not moved for several years may not be able to use it, and no one dares to move it. Based on Model, I understand that tumbling Bull Baby, of course, the bigger goal must be that we all have the same standard, the same dream, and the same perspective. Open openness has two meanings: 1) support heterogeneity: through the standard model statement, we accept different infrastructures of cloud, virtual machine, physical machine and container, which means that we can expand our platform indefinitely. 2) in the cloud native era, we can reuse all kinds of infrastructure to make small workshops and experience the feeling of the next five-star hotel. By reusing cloud vendors and the open source community, our platform can seamlessly enjoy the capabilities of the open source community. Next, let's take a look at how OAM achieves this goal.

2.1 Component

Component is a part of an application, which is usually described by the developer. The following parts can be described as a Component: 1. Research and development will package their own programs into an image and deploy 2. 5 through Deployment. The R & D statement requires the use of a 4-core 8G mysql5.7 database

These two descriptions are both component. In short, what R & D can describe can be called Component, because they are all part of Application. The Model of this part is mainly reflected in that through the standardization of Component, R & D can complete a definition of application from the perspective of R & D by paying attention to very few things that R & D needs to know.

At this level, I understand the ability of the infrastructure team to define applications for R & D. R & D only needs to declare the corresponding component components according to the needs of the application, without paying attention to the specific implementation of the underlying infrastructure.

2.2 Trait

Trait is the means to provide the service-oriented capability of operation and infrastructure. Operation and maintenance can add the corresponding Trait to the application according to the description of the application. What can be called a Trait? For example, elastic scaling may be a kind of Trait, log may be a kind of Trait, monitoring may also be a kind of Trait, and several actual Trait: 1. Research and development launch a java application, the operation and maintenance side according to the standardization and service capabilities, will automatically add the corresponding monitoring, which is actually the monitoring service capabilities 2. Apply grayscale release, find problems, and roll back actively, which is actually the embodiment of the service ability of the publishing system.

Through the output of operation and maintenance capabilities, R & D does not need to pay attention to the underlying operation and maintenance details, but only needs to declare the capabilities that the application wants to have, and can realize the automatic hosting of the underlying application operation and maintenance without paying attention to any underlying details. and all kinds of operation and maintenance capabilities can be combined freely to realize the stable and efficient operation of the application.

2.3 Policy

Policy is actually a concept in Component, which actually reflects the requirements of research and development. For example, if R & D proposes that the response of my application needs to be delayed within the 100ms, then the operation and maintenance staff can provide the corresponding Trait according to this Policy combined with their own service-oriented capabilities. In fact, R & D does not need to know how the operation and maintenance guarantee the SLA. It only needs to provide the R & D request Policy, and it can actually complete the transmission of the request. Everything can be described and measured.

R & D conveys the demands through the declaration of Policy, and the operation and maintenance staff combines the operation and maintenance capabilities according to the requirements to provide corresponding protection. Everything is digital, which is not only the embodiment of the service ability of operation and maintenance, but also a good bridge for R & D and operation and maintenance to trust each other. At the same time, R & D does not need to pay attention to all kinds of underlying details.

2.4 Application Scope

In the application boundary, we first need to understand the boundary, which mainly defines the boundary of a class of applications with some meaning. For example, in the public cloud environment, the boundary is usually divided according to the VPC network. Through a unified network configuration, multiple network areas can be divided, all of which belong to Application Scope. The more complex scenario is multiple data centers with quick stop losses. At present, most companies have multiple data centers for disaster preparedness. In fact, each data center will be divided into a boundary. If a center is found to be dead suddenly, that is, under the corresponding Application Scope.

And ApplicationScope can be combined arbitrarily, we can use this game to uniformly manage a certain type of application, pay attention to its corresponding state, and then combine our Trait to realize the construction of various scenarios.

2.5 Application Configuration

Above, we declare the component that makes up the application, add the Trait of operation and maintenance, and demarcate the boundaries of the corresponding network or application layer through AapplicationScope. These components are declared independently and can be evolved independently, realizing the standardization, modeling, loose coupling and free assembly of application access configuration. The remaining step is to combine Conponent, Trait, ApplicationScope and so on through Application Configuration. That is, our final application statement, based on this statement combined with the final state-oriented design, OAM will host each component separately according to the rules, and we can also reuse the excellent Component and Trait of the community to achieve rapid delivery of the platform.

Is it helpful for you to read the above content? If you want to know more about the relevant knowledge or read more related articles, please follow the industry information channel, thank you for your support.

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.