For new set-ups, Linux developments and test environments were Docker or Oracle Virtual Box my top choices. To test container systems like Docker or container orchestration like Kubernetes, it makes much sense to try it on Linux servers instead of Desktop Operation System. In Machine Learning development, is it best practice to set up a new Python or R environment as a container or virtual machine (VM).
Docker runs well on macOS or Windows for typically developer steps. To simulate production environments, it makes more sense to work with Docker and Kubernetes on a Linux Server. For a long time was Oracle VirtualBox the tool of my choice. It's a free and open-source hypervisor for virtualizing. It runs on all crucial host operating systems Linux, Windows, and macOS. To set up a Linux-Server, I need fifteen to thirty minutes for a new clean Ubuntu server. Three weeks ago, I tried to start Multipass. It provides to build straightforward Ubuntu VMs. Multipass also runs on Linux, Windows, and macOS. To set up a new version of the Ubuntu server costs only up to ten minutes, including installing all updates and upgrades and Docker as a container system.
Multipass allows at the launch of a VM to set CPU kernels, RAM, Disk space, and a Network to the host. It's also possible to mount directories from the host into the VM. Currently, Multipass does not provide commands for import or export virtual machines. For my use cases, it is no problem. If somebody thinks more about use cases in productive environments, it must be taken into account.
I am happy with Multipass; it closes the gap between Docker and VirtualBox.