Reading the comments, they seem sadly fairly black and white. Either you need to have pets on tin, or cattle via cloud, but that never was the case. I worked

Reading the comments, they seem sadly fairly black and white. Either you need to have pets on tin, or cattle via cloud, but that never was the case. I worked at a hosting company ~2007 whom was an early IaaS provider. We PXE booted xen nodes, that automatically connected to our management layer, allowing customers to provision virtual machines. Most of our own fleet would be cattle well before this was meme worthy. Today, you could bootstrap a k8s cluster with almost no effort on tin. You'll quickly have autoscaling cattle and a distributed cron. Sure you'd probably pet etcd and maybe the API servers. Running a database, API, and small management layer is well within the responsibilities of a professional system administrator. If this is beyond your orgs / teams capabilities you probably should use the cloud provider. P.S. Not having a team that can run production services without outsourcing the database is fine. We all have different specilisms. The storage layer is a bit more complex if you want to roll PVC. You shouldn't bootstrap a $1m team to defeat a $500k cloud bill.