Spark config: is there an advantage to running master and worker(s) on different VMs / machines?

Keywords´╝Ü apache-spark


I'm running a Spark cluster, with the standalone cluster manager, and I'm wondering whether I should be setting up a separate VM for the master. The alternative is to run the master on the same physical machine as one of my workers (background: I have 4 physical machines, and I have to decide whether I want to also request a VM from the infrastructure group. As part of that request, they will want some technical justification).

Any thoughts or (even better) experiences? The documentation seems silent on this question.