Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> BUT ALSO the care and feeding of Kubernetes clusters themselves is a significant time investment

AWS and GCP will do this for you for a small markup



While this is true (btw, Azure as well has a managed offering) the complexity is not the only issue. Cost!

The article is basically advocating, "if you have more than one service, you need K8S", and sticking your app on Kubernetes in the cloud (managed or not) is a huge cost upgrade from a single or handful of servers.


You're right, but the higher your service count gets, the more you should probably be on k8s. Large scale ECS or Lambda stuff is much worse to maintain


I would disagree in a couple of important ways:

1. The higher your service count gets, the more you should probably be on some container orchestrator. There is more than one. 2. The number of services you have to manage to justify moving to a container orchestrator is most certainly not just > 1. That is crazy. I don't know the right number, as it probably varies according to how mature your organization is and how well the parts work together. But it is definitely higher than > 1!

My last point is: If all teams worked together smoothly, or one team had infinite capacity, or if automation was perfect, etc. etc. etc. we wouldn't need containers, let alone container orchestration. These are solutions to people problems, not technology problems.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: