[chubot]

This looks pretty interesting, although I'll have to dig more into the examples to see why they chose this set of primitives (multipipes, multipipe blocks, and stored values).

Here is a 2009 paper, "Composing and executing parallel data-flow graphs with shell pipes", which is also a bash extension. (I'm impressed with anyone who successfully enhances bash's source code.)

Although it has a completely different model and I think more suitable for "big data".

https://scholar.google.com/scholar?cluster=98697598478714306...

http://dl.acm.org/citation.cfm?id=1645175

In this paper we extend the concept of shell pipes to incorporate forks, joins, cycles, and key-value aggregation.

I have a printout of this paper, but unfortunately it doesn't appear to be online :-(