While I get the whole "you can do it on one node without all the complexity" thing, I do still wonder if map-reduce-synchronizer + coreutils is better than the

While I get the whole "you can do it on one node without all the complexity" thing, I do still wonder if map-reduce-synchronizer + coreutils is better than the behemoths that are the distributed ETL platforms right now. All the system would need to do is make a data file available on a node and capture stdout of the unix pipeline. I know gnu parallel does some of this.