DevOps is a fusion of traditional practices

On May 6, 2017, Wang Jinyin delivered a speech on the integration of DevOps and tradition in “DevOps&SRE beyond traditional Operation and Maintenance”. IT big said as the exclusive video partner, by the organizers and speakers review authorized release.

Read the word count: 2786 | 7 minutes to read

Abstract

The overall framework of DevOps is first a collection of engineering practices, and second it focuses on overall business and lifecycle management. In addition, it is still based on management ideas, emphasizing automatic pull management.

ITIL is oriented to the whole management process, with priority given to specifications, low efficiency and high cost. DevOps is the automation of execution capabilities for IT operations.

ITIL features the management of offline tasks, but DevOps has invaded online services today.

According to our practice, in the process of traditional industry interaction, our products and ITIL products docking came to three modes.

• The first mode is the online service opening process.

• The second is the major change process.

• The third is the agile release process.

The first is the waterfall flow model. Then there is the agile r&d model. As new business forms emerge, emphasis is placed on the integration of end-to-end capabilities. This is the software development model of DevOps today.

As r & D models change, so does the ratio of work undertaken by each role, and R & D becomes more and more important.

Five ideas are emphasized:

1. Break islands through continuous service delivery value chains.

2. Integrate development and operation and maintenance capabilities to become a collaborative team.

3. End-to-end continuous delivery process transformation.

4. Accelerate and shorten the time to realize value for new applications and services.

5, does not affect security, compatibility and performance.

I’ve divided the DevOps system into six dimensions plus a culture. The six dimensions cover organization, process architecture, tools and infrastructure, and metrics.

Organization: First, barriers between organizations must be broken down, and second, DevOps organizations need a collaborative culture that is product-oriented, not project-oriented.

Process: The perfect combination of lightweight processes and automation tools ensures a high degree of enterprise agility; Automation first, process second.

Architecture: Architecture is divided into application architecture, infrastructure, and data architecture. Application architecture and infrastructure are fairly straightforward. Application architecture is about microservices architecture, and infrastructure is about standardization infrastructure.

Tools: take quality cost and efficiency into account at the same time, with practical.

Infrastructure: Virtualization to containerized platforms.

Measurement: Measurement system can continuously drive capability optimization.

In the vertical dimension, competency assessment is divided into five levels, which refer to CMI.

Quota defines the maximum number of resources that can be consumed by a namespace, and the BatchJobController calculates values that float back and forth.

Start small based on a character and a scene.

Perform imports based on a system or a functional domain.

Avoid coveting too much.

In order to narrow the scope, the name of CMDB will be changed to IT Resource Management Platform in the next release. After narrowing the scope, only infrastructure and application resources are managed. Infrastructure resources are assets. After the asset status is managed, the resources used are seen from the perspective of the application, and the two dimensions are strongly correlated, and then various management scenarios of the application are constructed on the upper layer of the application. By it to further drive the flow of CMDB, the application of this dimension is in line with the characteristics of high frequency. CMDN is the core metadata of IT operation platform.

Because there are many roles in operation and maintenance, and the final capability management process is also very many, it is necessary to add roles and scenarios to finally derive what kind of capability management platform should be built. Operation automation permeates every character and scene. Job and scheduling capabilities are low-level platformization capabilities that should be used by the upper subsystems.

Job management and scheduling management is a platform-level capability, which does not need to be understood in a scene, as long as it is considered that there is an atomized thing in the components of automation, and a scheduling engine is needed to arrange this atomized thing, there are two elements is enough. Based on the abilities of both, categorize for character and scene convergence.

Tools are a culture that allows us to build all kinds of atomic tools that piece together our capabilities and use them in various scenarios.

Good experiences must be automated, not documented. Tools are effective means to really drive change.

I thought it was going to be a form of virtual organization, so I made an improvement. The central role was removed and a split was made with technical thinking. But this one emphasizes the technology’s ability to build a continuously delivered, end-to-end DevOps platform through a continuously delivered platform team that is completely aligned with the technology and organizational structure. Do not change the existing organizational structure, through some means of technology to strengthen the organization’s mapping ability.

The core operating principle is to drive the value stream of the home customer.

Identify the entire value stream by looking at activities that are ineffective, activities that do not add value, and activities that add value to the customer.

The product must have a platform thinking. There is a boundary to this platformized thinking that cannot be crossed. At the same time to adapt to the status quo of each company, how to plug in some of the company’s existing capabilities.

From a DevOps perspective, IT must be the platform for the delivery pipeline, the END-TO-END scheduling of IT. The adaptation layer of the whole plug-in is the protocol standard of the plug-in, which puts the developers, testers and operation and maintenance capabilities into the interaction pipeline through the plug-in. The criteria for many plug-ins here can be thought of as job platform and scheduling platform capabilities.

Be sure to automate your own capabilities before you automate others, never cross the line. Is automation a process from point to line to surface, a process that gradually covers more roles, or a process that gradually covers the environment?

Continuous delivery is the best engineering practice of DevOps, based on the deployment pipeline with the goal of rapid delivery.

How to cover and merge the operation process after the application goes online? I divide it into several stages, including human and worker operation and maintenance, platform-based operation and maintenance, incubation operation and intelligent operation.

Operation and maintenance is also divided into the division of operation stage, the division of objectives, the coverage of the scene, the revenue achieved, and the resources needed.

Application is becoming more and more important today. The cloud forms the basic ability to serve, so how to abstract application into three dimensions, one is the dimension of resources, the other is the dimension of action, and the third is the state.

Applications must run on resources. An application change equals a resource change. The state of the application is equal to the state of the resource.

Finally, the evaluation and measurement of the entire state of the application is the evaluation and measurement of resources. Status is divided into intelligent monitoring, data analysis. Monitoring is problem oriented, data analysis is service oriented, data visualization.

Application is becoming more and more important today. The cloud forms the basic ability to serve, so how to abstract application into three dimensions, one is the dimension of resources, the other is the dimension of action, and the third is the state.

1. Change duration

2. Service recovery duration

3. Release frequency

4. Change failure rate

Reasonable indicators drive the construction of different organizations, high performance organizations, medium efficiency organizations and low performance organizations.

This is the end of my sharing, thank you!