For a master thesis in the field of distributed computing (think of hadoop and two-level schedulers like Mesos) I'm setting up various experiments on a university cluster.
However I'm already piling up bash scripts which function as a driver for the experiment. I miss composability and reuse between subparts of experiments.
Are there any frameworks out there which allows one to reuse various parts without ending up with tightly coupled components.
I've found a possible candidate called Nepi, but was wondering how other people bring some well defined form in their distributed experiments.
Here is a a good survey on experiment description languages.