To enable predictive maintenance, large amounts of data must be analyzed from a large number of sensors. However, data volumes of this size put a strain on data paths and storage. How could this problem be solved?