@simongray depending on your persistence approach, you could also consider named graphs for separating datasets