What Is AIOps and Why Should You Care About It?

Sylvia FronczakMon, 04/29/2019 - 15:20

Although we’re still just getting used to DevOps, there’s a new Ops kid on the block. And this kid picks up where DevOps left off, providing better operations and monitoring through increased data processing and automation. How? It does so by using machine learning to find new ways to automate operational tasks.

In this post, we’re going to take a look at AIOps and see what it has to offer us.

What Is AIOps?

AIOps stands for Artificial Intelligence for IT Operations. You may have previously heard it called Algorithmic IT Operations Analytics. The ideas behind both phrases embody the same mindset: we use algorithms or machine learning to better monitor and operate our application ecosystems.

Simply put, AIOps supports and automates operations based on data.

And the growing market of AIOps platforms provides easy access to all that data within your operational control. But more than just providing data, AIOps provides automated data analysis and automated responses to regular DevOps operations.

Still not clear on what all that means? Well, to get a better understanding of AIOps, let’s compare it to DevOps and see what else AIOps can offer us.

How Is AIOps Similar to DevOps?

Both AIOps and DevOps require a culture change where we look at systems from a holistic view. Therefore, they both encourage creating automated processes and monitoring that encompasses the whole ecosystem of the application.

They both go beyond the silos of products, technologies, and architectural layers. And they both focus heavily on automation.

Furthermore, AIOps and DevOps both automate repetitive tasks and work to increase collaboration between teams by giving each a common goal and language. And they both use data to identify problems and vulnerabilities.

However, there’s a difference between where the automation of DevOps ends and where that of AIOps begins. Let’s look at that further.

How Is AIOps Different From DevOps?

Whereas DevOps automates the path from development to production with simple readiness checks and automated tests, AIOps takes us a step further to use additional data. AIOps comes in not just to automate the path from development to production, but also to use data to predict the effects of deployments and regular operations. It also responds to changes in the production environment automatically based on metrics and data.

As another example, let’s look at CI/CD. When implementing CI/CD, DevOps wires up the plumbing and tests that you’ve written to ensure a smooth deployment between environments. Alternatively, AIOps uses data from each environment to enhance the rollout in the next environment. Additionally, it works to predict performance and outcomes based on the past experience of not just this rollout, but previous rollouts as well.

Furthermore, while DevOps still depends on manual intervention to investigate root causes, AIOps uses data to predict performance issues, suggest optimizations, and even process root cause analysis. It could even automate changes based on the data provided.

So AIOps continues where DevOps left off. It provides more automation by using even more data than we currently do.

Why Do I Need AIOps?

Reading this, you might think it all seems a bit pie-in-the-sky. But we’re definitely getting to a point where AIOps can make a real difference in our daily operations.

As our systems become more complex and distributed, it also becomes more difficult to manage all of the different technologies and infrastructures. And with that distribution, we get more and more data where AI and machine learning can make that difference between manual monitoring of metrics and a system that monitors itself.

Let’s look at some key ways in which AIOps attempts to get us further.

Reduce Human Intervention

DevOps has gotten us part of the way to automated systems by automating repeatable tasks and allowing us to recreate infrastructure through scripts. However, we still heavily rely on humans to look at logs, alerts, and metrics to find issues.

AIOps is there to bridge the next gap.

You see, with human intervention comes human error. Therefore, we can easily overlook issues because we don’t consider certain data points or miscalculate the time at which an incident occurred in different systems.

But using AI and machine learning can help identify patterns and problems before we even realize they’re there. And we can further reduce the need for humans to investigate logs and uncover data that AI can readily process.

And why does AIOps have an edge over humans analyzing the data? Because AI can take in all the data. When humans review dashboards and logs, they can’t always absorb the entire volume of data. They leave things out. They focus on what those particular humans on that particular day think are important.

But AI doesn’t have to. AI can take in all the data and find patterns and root causes that would take us as humans much longer to find.

Provide Holistic Analysis

A common problem for many companies involves narrowing down issues that affect multiple services. For example, if you have a large distributed system with 20 microservices, 10 databases, and another 10 third-party applications that deliver streaming or data storing, finding root causes can be difficult. Additionally, it’s often hard to find out what problems are related.

With AIOps, you can have systems that monitor and correlate data across multiple systems and platforms.

Also, AIOps could correlate performance issues with code issues. Given enough data, AIOps can recommend changes in your applications that avoid performance issues or improve throughput. Actually, why stop there? AIOps, when used properly, can analyze the code, make changes, and push and test them based on data and static code analysis. Why not?

Improve Automated Testing

Let’s look at another place where AIOps can help. One problem with large applications involves the giant test suite. The tests take a while to run and might not cover relevant areas of the code. AIOps pushes the boundaries and runs only the tests that are necessary for a particular code change. Additionally, AIOps could alert us to code changes that could result in failures we’re not testing for.

Update Runbooks Automatically

This may be my favorite part of AIOps. We all know and love runbooks. They help to point us in the right direction when issues crop up in production. But there’s a dark side to runbooks—they require manual updates by developers who are pressed for time and focused on too many other concerns.

So what if our machine learning could detect required changes in runbooks based on the code that’s being checked in? What if we could automatically update runbooks on the fly?

Sure, maybe not everything would be updated quickly, and yes, we’d need a review before pushing it out. But how awesome would it be to not have to worry about out-of-date runbooks again?

Fix Code Vulnerabilities Immediately

With DevOps, we use automated tests and static code analysis. But with AIOps, we have more tools at our disposal. In addition to identifying issues in our code, AI can be used to fix our code when issues arise.

Additionally, when vulnerabilities surface in our dependencies, AI can automatically upgrade to the latest version, run regression testing, and deploy our code without human intervention.

We might still be a ways off from all these automated processes, but we’re at the point where we should start considering them. We could soon arrive at a time when repetitive and easy tasks are left to our AI, leaving humans to solve the difficult business problems instead.

So What’s Next?

It’s time to investigate AIOps providers and see what they can deliver. Although the futuristic vision of systems monitoring and acting upon themselves might feel like it’s still a ways off, this field appears to be headed in the right direction.

And we should become aware of AIOps’s progress and promises to see how they can help our organization take on more operational responsibility and reliability with less human intervention.