HashiCorp Nomad vs Kubernetes - Comparing Complexity

Andy Davies is a software developer at Reaktor, a software consultancy. In his talks at Codemotion Rome 2019 and ConfigMgmtCamp 2020, he tried to temper some of the hype around Kubernetes, specifically arguing that it wasn't the best fit for every orchestration problem and that it adds a lot of operational complexity to an organization.

In fact, Davies suggests that you should almost never try to deploy Kubernetes yourself—get a cloud provider to do it for you. "You're reselling Kubernetes as a service" if you try to roll your own setup, he says. "It's a waste of your company's money."

His talk, titled "Nomad: Kubernetes without the Complexity" dives into several demos of an alternative cluster scheduling solution called HashiCorp Nomad. Instead of trying to solve all of the things that Kubernetes tries to solve:

• Service discovery
• Load balancing
• Configuration management
• Secret storage
• Feature gates
• Routing
• Storage orchestration
• Autoscaling
• Container management
• Rollouts & rollbacks

Nomad just focuses on scheduling. It does have some storage orchestration capabilities but mainly it is built to be a bridge between orchestrating newer container-based infrastructures and older VM, binary, or other workload-based infrastructures. In one Nomad cluster, you can run Docker containers, VMs, and Java JARs—as an example. Kubernetes, on the other hand, is mostly built for containers only.

In this talk, Davies will demo several features of Nomad to showcase its capabilities. He'll demonstrate how to set up a Nomad cluster, deploy containers and applications to it, and then integrate it with existing services and infrastructure.

Here's the GitHub Repo for his demo. And here are his slides. (Use the arrow keys)