Yeah. I think there are still use cases like processing TB of click data quickly, but the original assumptions about how much RAM/disk/network machines have has changed since the inception of Spark.