[bunderbunder]

I'd constrain the description of what it's good at even further. The kind of data processing parallelization it's designed to support is map/reduce. That's only one out of a wide array of possible ways to organize a distributed processing job, and not all tasks are well-suited to it.