Instead, consider Map/Reduce. It has a bunch of the 15 properties, and works beautifully:
Levels of scale – small computations, the big aggregator in the center.
Strong centers – each computation is a center on its own.
Ambiguity – You can't tell which sub-computation is which; in the end it doesn't matter.
Roughness – each computational node can have different speeds or capacities.
Not-separateness – no computation makes sense without the others.