Interesting project. Has anyone tried adopting something like this in their database cluster? I would like to know how it performs in practice. Is it useful?