Continuous Delivery Foundation cdCon 2021, 6 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Ten Core Strategies of DevOps - Forest Jing, DAOPS Foundation

Description

Ten Core Strategies of DevOps - Forest Jing, DAOPS Foundation
Speakers: Forest Jing
When I ask if you are practice DevOps? Most of you will say "Yes". But how do you practice DevOps, if there are any strategies you could follow in the whole companies? This topic will share my experiences with my team in last few years in practice DevOps in different kind of companies. I will share with you : 1. Fine-grained low coupling 2. Continuous flow in small batches 3. Comprehensive means to ensure quality and security 4. Automation and self-service 5. Speed up various activities 6. Fix in time 7. Complete record and full display 8. Coordinate to complete the overall function 9. Ensure consistency 10. Continuous improvement based on measurement

For more Continuous Delivery Foundation content, check out our blog: https://cd.foundation/blog/

A

Hello, citi fans welcome to city count, I'm forrest the chief evangelist from dubs foundation. Today I want to present a topic about devops. The name of this topic is turn code strategies on devops, uh as you know why we just want to introduce devops and another messenger like their ideology.

A

We just want to make business success, yeah their revenue, their profit, more users. So I want to split the building success into two parts. The first part is about product discovery and, second, part is about product delivery. We need to find more useful requirements or functions for our users, but we also need to deliberate it in more efficient way, yeah. So that's a different uh different rule to make their business success.

A

So I want to make clear about the develop school just forward more faster, better and norm. So I want to share some detail with you uh just in develops. I just want to make it clear that most of their development is focused on to solve their problems about their product, realization sites. So we just focus on their development, the test and the operation security we want to just to provide their product and service to realize the value. So where is the value? Value is from the first part about product definition.

A

This part is most focused on to design the product like we will introduce the early startup to just to make their requirements is fit for their real real requirements yeah. So this forward is focused on the flash faster. We need to response. The requirement from our customer just faster, and we need also do more requirements or features requires productivity, and we also need a better quality and security, and your cost basically could be an internal cost.

A

So let me think that's their forward could conclude, could be your conclusion about your google of developments, because just their basic uh just a it's a basic word, and we all make the blogs to understand what we are doing. So the most important thing is to establish an end-to-end development and operation integrated pipeline centered on flow unit and deployment unit. So what does this synthesize means, just maybe in the manufacturing factory, mainly just uh to uh just to build their car?

A

We need their pipeline so just in software, and we also need to need a pipeline to deliver value from users to their just like the development team to their testing to operation team and other unreal back to our user stand. So there is two just a concept. The first one is about the flow unit. What is flow unit for future regular storage or flow units? We will delivery just by a draw unit. By throwing on it, we will not just deliver a worry, just a big batch of requirements.

A

We want to just make it flow more fast, make it faster and about the deployment unit. It's not like a micro service. We do not want to deliver a whole software to our others. We want to adjust the delivery by their just a microservice just like this, so this is like the factory of tesla.

A

As you know, we will build it like this, so I'll just come to their 10 core strategy to share how we could to make it like a platform to deliveries or our features and we're just very fast, faster and could deliver more features. The first strategy is about small size and low coupling.

A

Most of people could know their high consequences and low coupling is just their first principle when they design or just software, but it develops it's not only the software architecture and we also need a small size and a low coupling organization structure, and we also need small size and low coupling software development delivery process.

A

I will shoot like this about their organization structure, just in traditionally they have their business department or their dba department, their test department and their edge operation department. So center solo, as you know, so we want to uh change it. uh Just there is a book named uh team topologies as to save it into four different teams. Tips I have just make summarize is like their infrastructure team. It's about their platform, it's a platform team.

A

Maybe they could build their developed platform and their enable team means to like their devops goals, so just some like kendall coach to a teacher team to master some skills, and the second part is about the application team. Let me just build their application. We are just like their stream and team to just like build their crm system and they're about to complicate some system team inside like their ai or big data team.

A

This is a organizing organizational structure where I just want to split it from like this to like this one and they're about the software architecture. This is the most familiar for you. uh We are not to just introduce more details and for the software development delivery process. Most of the processes like this, we are just an instrument yeah, so we want to make the process like this. No coupling or just a small size just from the beginning to the end, the beginning is circled and the end is just their service online.

A

So we want to make every unit every service could have a pipeline like this to deliver every feature on this pipeline. To our owner's hand, this is about their small size and low coupling.

A

And the second strategy is continuous flow in small basis, when we want to just change the process from the waterfall to scrum and right now, we want to focus on the array by feature. That is just. We have just noticed that if we want to deteriorate your just uh your big bags, yeah, not bad, then we will wait. You know wait uh not for people to integrate our call together yeah. So this is our processor method we need. We need to split the requirements we want to uh just to split into some small features.

A

We could describe it in various stories themselves and from their testing thinking some measure from their tps. We will focus on their own, peaceful and limited wip. It's our practice from their campaign message. It's also one to make it uh continue to flow, and when we want to introduce continuous integration and continuous delivery, it's also a very important practice and to make it like a float or continuous flow, and another very important practice we could to make its our company continuously is to remove the release window.

A

What is this when you really sprinkle it so maybe like? uh We could only release one month, one time- seven months, maybe like just last friday this month and we have a chance to release our software to online, so the management has only this time. So we need to wait for this time. So when that is, the release window is readable in a password, because we do not uh to make some like uh another change to their to their production environment, because every change in their production environment may be caused some incident.

A

So that is our second strategy.

A

A third comprehensive means to ensure quality and security. As you know, in devops, the current ending security is very important like we need to build the security and the quality in the process, the comprehensive assurance at all state.

A

It means we need to consider the quality and the security from requirement to production like we need to consider security of the requirements, and we need to do the security test requirement test and their api tests and the ui tests, and we also need to do something like their chaos, engineering in production, to do some monitoring and to make sure the production is already stable.

A

So we have a lot of types of testing, and another very important way we need to follow is that we need to fix production. You should immediately yeah. This is a very important principle and about some different tests uh like we have the manual testing and the automatic testing. I I do not think in device. We will just remove manual testing. Meditation is also very important, like in their uh euler certain tests, and sometimes we need to just test our software enhance you like some.

A

um If it's very easy to use yeah, it may be like the operating steps and we have the code level testing testing in in test environment. It means uh just like the early dentist and we're also about testing production, as we said, and another one is the build and deployment test too.

A

If the software could not be built, could not be deployment, and it is also uh just a quality problem and about your comprehensive use, it means the tips and the expensive tests. We also need to split it and we need to spread shift, lift and the builder quality in, and we also need to do where ships quality right. So that is a comprehensive means.

A

The first one is about automation and the self-help yeah. Maybe like a self-service, is not um the benefit of automation. It could do to save their time and cost less. You know, and we could make it repeatable when we, when we just deploy just uh in hand if it were means and every people will deploy it differently.

A

So we need to learn automation to make it repeatable, and it will also be real document means, and if we do it, just automation wait, and there are some logs and different records that we could to be audited and automation, a single priority in activities, and we also need to automate your process and just like the build we need to ultimately to build automation tests and about your process means that we need to apply yes, they're continuously era pipeline to make all the process.

A

It's also automation way, and another thing is also important in the processes- means maybe like the approval process. Whether leaders want to approve some, maybe just like their release or deployment.

A

We also need to automate it maybe like, even if there is a checklist when we find the technician really automation and we could to or approve the process automatically and another one is since automation is not only just, then we are not only need the automation. We also need. A self service have noticed that a lot of devops platform is not very easy to use, because there needs some expert to help their developer. To use it it's difficult, so we need also their platform their tools to be much more self-service and their another.

A

One is about your stability. Obviously it's also very important just then, if uh maybe you could get them to if they're developing the device platform, uh just a song, every developer could not to their do their development to wrap their code to push their code and to just to pack it to the coin. So there are two of devices, also very important to make it stable and about their fifth strategy is about accelerate air reactions.

A

We need to how to make uh it possible. We could do more features to make the productivity just to make it just higher that we need to accelerate every actions in the process.

A

Mostly just the easiest way is to use a hardware to uh to optimize the uh spin, maybe like we will introduce some good hardware to to compile. You know to figure test. So that's easiest way, but we'll be a software engineer, so we will have some different way. The parallel is very important in tests.

A

We need to do the test parallel, and that is a very important method strategy to do it, and we also need to avoid the duplication like we will just uh to maybe like to clone the code uh several times when we just in pipeline is other smart way, and we also could focus on that.

A

We focus only on increments what it means when we have a very big repository to store our code when we just clone the code, it will be just very small, so we could now we could replace the git clone command to the defense command and let me use scan record, uh you know, horror repository, it could be very small and we also could just to scan their increments and the next one is about your circuits.

A

Just uh maybe, eventually, probably who user may have encountered just a minute. We have a local repository to gesture 2k to catch their uh just my like some dependency, just in our local, so we just convert it more fast when we want to compile, they do not need to download the dependency from the network from their central uh memory laboratory. So then it's about their accelerator reaction to make it more fast in their pipeline process.

A

The sixth one means they're: they need to fix immediately. What does that mean?

A

Most of people maybe know that in see, maybe in their ci practice, yeah, we need to fix their or just uh students, and just maybe closer uh are ci like a raid. We need to fix immediately so this is also very important in your whole devops process, uh not only for the test test stages and also when we just release it online. We need to fix every problems, how to employ implement it.

A

If you want to fix their issues or incidents immediately, the first one first step. We need to make the notification as fast as possible, but there is a problem. We need to do the notification and when we just get into notification, we need to make it as a top priority to fix it, and just if we have just a artifact or we just consider uh just the field, and if there is our issue in their production, it's a high priority to fix it.

A

When you found me, you found it's very difficult to repair your issue to make it it's just to make it it could work. uh I think there is a relatively difference and the rollback is better than the pair well, you can't fix it in time.

A

We could to look back, uh maybe like just uh really just uh put your code into your trunk or your cap custom problem and you could to go back to launch to influence the other developer, and we also need to notice some detailed information like build logs and some just production locks to their people who will fix their fix their issue, make it easy troubleshooting yeah. So that is six yeah. It's a six strategy and for the seventh same strategy, is completely record and fully display in their developed courses. We need to record our pattern.

A

Like their feature user story, their deep, fake and their instant in production, we could use their sperm backlog to list all their facial. Your stories, we could use their command port to make them a.

A

Members could see their whole process their situation and we also to record the information about their pipeline and we need to use the version control system to record their tune. Their team set up their code and uh not only their code and like just like their test case, maybe like the passing code to render their test case or we will just to config in our web system, and they all need to have to need your version control and about that artifact.

A

When we just build a tesla cr package, we also need to store a thing like the nexus or not manufacturing system to also version control it and to save like some metal data, and that's why it's about the management of supply chain, especially in the div, stick out in. We think it's very important project manager or surprising. We need to know all just the packaging we uh depend on, it's really important, not only in the in general, and then.

A

We also think it's very important to management their air striking their ethics is to ensure count, assistance, consistency.

A

When we need to make it, we could make it we could make it repeatable if uh just like airy, uh maybe just like, for example, only use a docker, normally use a dock file to write all other uh to describe our environment. We could to build it uh to just build your docker image every bit more. It's also the same, and it's also standardization social standards. Every in just every image.

A

Every container is the same yeah so how we could to import it first, one is to make a standardization not only for the next dark skills, maybe when they use their answer or just pockets or stacks. It also could to use your configuration, your infrastructure as code to make it standard yeah, and uh we also need version control to management it could be just like russian is controlled by the system. If you want to change, you need some like your usual tickets, you could change it.

A

The next one is your coordinated to company to the overall function. From our view, we could now to split your team into very just like a different team than they could deliver only by themselves. There will be some just like their cognition come out, so they have different parts. We could to coordinate the first. One is their active texture level. We could define our our api, and so then we could to coordinate, and we we could know all their hiv will depend, and we also produce like the swagger to manage an ancillary high.

A

Then another important part is about their credibility. We need to give their conversion to our apis so that it could become critical and about your management level. If we have a non-subscribe team, we could use like a scramble scrum another measure like some some single ideometer like the safe or less or like discipline, the ideal development. We have another measure to management like this table and further engineering level. We feature involved in to multiple to employment units and also integrate and release much multiple depending on it.

A

What uh we could not all avoid their dependencies, so we will just like use the uh pipeline like a component pipeline and the integration component and use it like this way to make it together to deploy and for the commercial level we need to close collaboration in different teams, but I think there is a collaboration is about collaboration, yeah, international environment, driven continuousing improvements and just in different kind of measures or books. You could finally like the last step last week in davao hamburglar surveys.

A

The last way is to continue learning and experiment instrumentalization, and that's now, the conference for our agile development. It's also about in just due to improvement and lots of their family's developing thinking to seek perfection, and their last of their experiences of continuous theory is to continuous improvement.

A

So every measure, just as a framework, told us to improvements that, from our opinion, we need it, but we not only need either continuous integration. We need to use the measurements to join your continuous improvements like we will use some different metrics, like the current frequency. The need time is just like there are just their fair fit about their deployments. We use like these metrics to drive and to drive us to improve, and we also to use some management today like interested in scrum. We all have their adjuster to perspective. They do we chose meeting.

A

We could to review what we have done in this scrum, how we could improve it so there their continuous improvement is also very important, uh passing mr last uh strategy.

A

So that is all the strategy I have just summarized and my way practice evolves uh how my views and the strategy could guide us how to choose a practice and why we need this practice just like why we need second bank, I mean it's just a ci cd and just why we need to split our system into different microservices, so that is, there are some strategies I have made her summarizing this picture from their business from left to right. It's from their dna success to your google observer develops and we all found return strategies.

A

Every strategy have different kind of practice from their application architecture and ideology. It's mostly focused on management and the continuous delivery is most focused on their engineering practices and their si is focused on their at the operation. Side and folder are just tools and organization and security, so you can use this picture to find out if you want to. Maybe you want to just deer in small size and make it more coupling.

A

We are fun and we need a couple architecture we we prefer to derive by feature yeah and if you want to like a small science team yeah, their team should to manage themselves and like the continuous in small batteries. Most are lots of practice on their continuous theory. We need theirs. Yeah we need sci too, is to verify if their future is good, to rearrange their next step. So this picture, you know, detail and the relationship between your strategy to your practice.

A

So that is a reference, uh as you know, I'm just coming from the jobs foundation, so I will just share some um just uh just what we are doing some interesting things.

A

As you know, the center of software engineering, uh neither close or academic uh yeah or not transfers, so we're just trying to figure out. Why can't we develop standards like their open source software, like their civic foundation? Yeah we use a lot of people, will particularly use their open source to develop right things, techniques and sections.

A

So we think we also have to use open society to develop standards. So we have. We will just to set up a non-profit organization named observation in singapore.

A

Its mission is to commit into accelerating global digital transformation, so taking standards, we are not doing their software, but we are doing their standards yeah, their favor flavor digitalization, their ai companies practice and their security. It's just a compound learning jumps yeah. So, let's use their dumb foundation.

A

Let's say foundation we have made are just open source standards for their advanced software engineering models and it's a model just we describe it as a non-native oriented inter intelligent software engineer model, you could find it. It's just a whole process from the development management to your software delivery to a technical operation, and it's only it's not only the process, but also the architecture and the security and the organization and platform.

A

It's just a big feature of their of this model, uh especially in develops with difficult just. We think it's very, very important right now. We do not want only user devops to deliver it faster and more. We need to deliver in the secure application, how we could delivery secure application. We need a security process, so that is we focus on other dipsy cups. It's also a very important part. In their essence, we also have founder a very way called mth. It's amazing to practice devops based on their essence.

A

The first one is about the model. I think every big company needs internal mode or internal standards to just to guide your guide for their feature, team or devops team to just how to practice roles, and we also need your tools to support and models from my viewers, and we think we need to adopt more open source in their developed, developed platforms and develop platforms.

A

Like live, entertainment, xxx.

A

The construction of your claw is anterior cell series, as we mentioned before, and not only there is more than two hours if we want to uh just to practice or develop more successfully.

A

We need to human human is maybe like the devops codes they could to uh just uh to adoption in their model and the trolls you're just like who uh I'm also the ambassador of the devops institute, and uh we also need to build her system and to for different kind of engineers like their sre, like uh the gypsy clouds and the different teams, the goodie masters and skills about devops yeah, so that is sir and she actually at last.

A

I just want to share with you all three sentences, um I'm just uh giving you an another thinking. The first one is under critical paths of our future from development to production. In addition to functioning development, the area stays as short as possible. We could keep their tests as short and keep their test stage as short as possible. They could fix their deployment release date as short as possible. We need more time for development.

A

The second one do not change the existing process and people's behavioral habits. Improvement develops its own realistic in itself. We need to change the process. We need to change the behavior of our teams and several your mechanisms. Change behavior behavior improves culture. So that's my presentation today thanks a lot for your listening and I'm very, very happy to bring in silicon, I also, hopefully, could do to see each other in your future. Facebook is to face in the future. Yeah. Okay, thank you. That's my topic today, hello there.

A

If you have any questions about strategy or devops, you could to put your questions in your q, a and I'll be very lucky to answer any answer. Any questions.

A

uh Yes, you could have the slide yeah.

A

You have checked the description below then to the pdf of the uh presentation slide you could find where their following link just below and if you just want to um to just uh to have the pom-pom file, maybe I think so yeah and you could just find me uh just uh only on linking you- could find your forestry. You could find me and uh we could to share yeah. I could share your powerpoint file.

A

Yeah amy has just an user link yeah the you could found your slide.