Ingesting data and evaluating it according to some model, sounds like what a good cluster node would do. No need to permanently store neither the data nor the output on the node itself, but you might want it to update its process via learning.