Chapter 7: Understanding the Impact of AI on DevOps

In this chapter, we will introduce artificial intelligence (AI) and what the impact of AI is on DevOps. We will discuss how this is driving a shift left in operations, by enabling the fast identification of issues already at the beginning of the DevOps cycle, using AI and machine learning (ML). Before we can implement systems such as AIOps, we need to get the enterprise ready for AIOps in the first place by creating visibility of all IT assets and workflows and mapping them to AI-driven processes. Next, we need an integrated toolset for both development and operations. Leading public cloud providers offer native toolsets, as we will see in this chapter.

After completing this chapter, you will have a good understanding of the concept of AI in DevOps processes. You will also have learned how AI-driven systems can help in achieving shift left. Before we discuss the possible outcomes and benefits of AIOps, we need to create full visibility...

Introducing AI and ML

In this section, we will briefly introduce the concepts of AI and ML. There have been complete bookstores worth of books written about AI and ML, but in this section, we will merely give a definition and describe how these concepts will change development and operations:

• AI: The broadest definition of AI is a computer technology that simulates human behavior. In most cases, AI is used to express the fact that software is able to react to events in an autonomous, intelligent way by deducting and analyzing and, by doing that, reaching decisions without human interference.
• ML: After AI is machines that learn how to perform tasks and execute actions by analyzing earlier events, and then use this experience to improve autonomous decision making. To enable this, both AI and ML as technology need data and they need to understand how to interpret this data.

AI and ML are not magic. You will need to define the scope for these technologies, just as...

Understanding the shift-left movement in DevOps

Shift left has become a popular term over the past years. But what do we mean by this? It's about moving activities that were originally planned at a later stage up to the beginning of a process. This is typically the case with testing, which for a long time was executed as soon as the whole product was delivered to a test team. Shift-left testing has become an important paradigm in DevOps: executing tests as early as possible. By having tests already from the beginning of development, issues will be found much sooner and can be fixed in that early stage. It will improve the end product. The following figure shows the impact of shift-left testing:

Figure 7.1 – Impact of shift-left testing

The shift-left principle can be applied to more processes in DevOps. Think of the very first step in DevOps: design. IT teams, both software developers and cloud engineers working on the infrastructure, should have...

Defining the first step – DevOps as a service

Consistency is the key to success. That applies to almost anything and it certainly applies to DevOps. Dev and ops need to collaborate in the same toolset: that is what DevOps as a service is about. DevOps as a service enables shifting left, but is also a good starting point for implementing overarching monitoring systems, including AIOps.

Note

AIOps is way more than just a monitoring tool, as we will find out in the following chapters. However, AIOps starts with the monitoring of complex environments. By gathering data from these systems and analyzing this, it will be able to track and remediate systems and processes, including the automation of repetitive tasks. AIOps is capable of discovering patterns for which it can define automated triggers. But it can't do this if it can't monitor the source systems.

DevOps as a service will track every step in the development and delivery process, but the real value is...

Creating the IT asset visibility map

There's a famous line in Alice in Wonderland: "If you don't know where you are going, any road will get you there." You can actually turn this around: if you want to go somewhere, you need to know where you're coming from. Let's put this into practice: if we want to transform the enterprise, we need to know what it is we are transforming. That's why every approach to digital transformation starts with assessments and discovery. An enterprise needs to have full visibility of all of its assets. The following figure shows the basic steps in a migration and transformation plan, starting with the assessment:

Figure 7.3 – High-level plan for migration

When all assets have been identified, we can start the planning of migrating and transforming these assets to a new target landing zone, typically a platform in the public cloud. Applications need to be validated so that the right strategy...

Measuring the business outcomes of AIOps

In the previous sections, we discussed shift left and saw how we can define DevOps as a service. Next, we learned how to create total visibility of all assets in the enterprise as a starting point to implement AI-driven processes that will help improve development, deployment, and operations and with that, accelerate a shift left in IT. How will AI help with that?

• AI is about analyzing data. AIOps is no different: it analyzes operational data and is able to give recommendations on improving systems in terms of performance and efficiency.
• To get valid data and recommendations, AIOps has to reduce noise. Noise is a very common problem in operations and specifically in monitoring systems and CMDBs, as we learned in the previous section. What is really an issue and what is a false alert? AIOps is capable of analyzing these alerts and, with the help of algorithms that group alerts, can identify and prioritize these. The outcome is that...

Summary

After a short introduction to AI and ML, this chapter discussed how these technologies will help in making better software and more reliable systems. AI enables the shift-left movement: shifting things that were typically done in a later stage to the beginning of the development and deployment cycle. With AI, it's possible to detect issues in a very early stage and by means of automation, AI will also be able to trigger correcting actions.

Since AI and ML are learning systems, they will learn how to predict and possibly prevent issues from happening. For this, AI needs real-time data coming from source systems, hence the first step is to get a total overview of all assets in our IT environments and make sure that these systems are monitored, providing real-time logs. We learned how to create this full visibility using five layers.

In the last section, we discussed KPIs used to measure the outcomes of AI-driven systems. Although AIOps is still relatively new, the...

Questions

1. Design thinking is a method to create a shift-left movement. Design thinking starts with evaluating the perspectives of all parties involved in the development. What is the term that is used to describe this step in the methodology?
2. AWS offers DevOps as a service using native tools. What are the three tools for building the code, planning the deployment scenarios, and the actual deployment to production instances?
3. What does MTTA stand for?

Further reading

• AI Crash Course, by Hadelin de Ponteves, Packt Publishing, 2019
• Blog by Clive Longbottom: https://searchitoperations.techtarget.com/definition/DevOps-as-a-service-DaaS
• Azure DevOps Explained, by Sjoukje Zaal, Stefano Demiliani, and Amit Malik, Packt Publishing, 2020