aitap/nodepool

Dynamically distribute tasks between R computational nodes (mirror of a Fossil repo hosted elsewhere)

nodepool's Issues

Clearer language in Readme?

So right now there is "pool server", "pool machine" , which I think are equivalent, and then there are "nodes" and "clients", and it's not 100% clear the relationship between them.

I think from the description that a "client" machine is where you have your task that you actually want to run, and it connects to and submits jobs to the pool. The pool server is managing everything.

For people used to cluster computing, these terms are probably second nature. But if you've not used a lot of cluster compute, then it's not easy to parse. A diagram, or a link to another resource (maybe the parallel package docs) would probably help a lot.

I like this idea, being able to setup a cluster where if a member dies or is taken offline then the job just gets resubmitted to another is really, really nice. Although the setup is a little harder than for parallel::makeCluster, it looks incredibly useful.