Benefits of kubernetes

This content is translated by Notion AI. Original content is written by a Korean.   

After deploying to a non-Kubernetes environment, I realized that Kubernetes has many advantages.

You already have a Controller and Scheduler.

When I first studied Kubernetes and applied it to my work, the following points stood out to me. Over time, I've come to take them for granted.

• All you need is YAML, and it deploys itself.
• Pods restart themselves if they are in an unhealthy state!

When I transitioned to a non-Kubernetes environment, I initially thought it wasn't as obvious. I developed my own controller for deployment, which involved creating logic to detect and act on unhealthy applications. Fortunately, AWS doesn't provide controllers and schedulers like Kubernetes does, but I would have spent a lot of time developing and optimizing them myself in a non-Kubernetes environment.

• Kubernetes is all when Kubernetes is installed

The second thing I've noticed about Kubernetes is that it just works.

I thought about what I meant by "Kubernetes is all you need". It's because Kubernetes is easy to operate in both hybrid and multi-cloud environments. You can literally go into any environment, whether it's on-premises or in the cloud. As long as you have Kubernetes installed, you can deploy and operate the way you've continuously operated. From an operational standpoint, the absence of environmental considerations is a significant advantage. From an infrastructure perspective, of course, it's a different story, but from a platform perspective, Kubernetes is a huge advantage.

I recently met with a team from another company that has designed its operations around a Kubernetes environment. They've been designing and building their MLOps system since 2022, guided by the design principle that "All we need is Kubernetes to make our system run." The reason they created this design principle was to ensure the system they built would run on-premises.

The reason for this design principle is that we have a mix of on-premises and cloud environments, and it would take a lot of time to design for each environment, which we don't want to invest in, so we created a principle of "all we need is Kubernetes to run." Fortunately, in 2025, many AI applications are optimized for Kubernetes, so it was a good choice.