Kubernetes (also known as K8S) is an orchestrating engine that operates over a pool of machines—physical or virtual—and can spawn and wind off containers dynamically. Technically, any container technology can be used, but Docker is by far the most popular one. 

For example, it is typical to use Dask on Kubernetes—in this case, Kubernetes will spawn more worker machines when you need them, and can either shut them down or switch to other users, once you're done—all without your intervention. Similarly, it can preserve a composition of containers performing different roles—for example, a load balancer for a web API, which then will redirect requests to different workers, who might operate on one or many database servers—all under the control and orchestration of a Kubernetes server.

This may seem like an approach that's too large and complex for a beginner developer and small services, but it is way easier to operate your services that way: you won't need to log and manually set up specific containers. Many technologies, including load balancers, databases, and so on, have pre-generated images already. In fact, you can even find and reuse predefined instructions on how to start running systems as a whole—including multiple servers—by default.

For more context on this technology, check out this video on Using Kubernetes for Machine Learning Model Deployment, by Niels Zeilemaker, PyData Amsterdam, 2017: https://www.youtube.com/watch?v=f3I0izerPvc.

Being a developer means constantly learning new things. I can guarantee that, next year, there will be at least a couple of new, cool packages that every developer should learn about. New technologies usually boost productivity and streamline the development process. The secret to productive learning is understanding the scope and requirements of what you actually need—not jumping on a new cool thing just because it is cool.