[infinet]

Very interesting!

For completeness, the job in the article is not strictly parallel, instead, they are separable tasks, which fall under high-throughput computing. HTCondor is a popular choice and is widely used in physics. If the cluster has a shared filesystem such as NFS, it does not even need ssh to each node. Moreover, nodes can share the same R installation on NFS, remove the complexity of installing same set of R packages on all nodes. HTCondor is available on Ubuntu.