@rorcde With Spark you also need to manage data layout (partitions) pretty well to have good runtime. Something closer to a "real" DB may even have better performance. When Spark started, scalable DBs weren't that common, now there are many options.