Darren, I have an overall flow with sub-tasks that all pass data to each other, read and write .csv files to Google Cloud storage (GCS), plus read and write from a database. It all depends on what I need for the data. If I just need the data to pull right into another subsequent task I just pass it in. If I need to archive the data then I might put the data into GCS or a database. If I need to access the data with another application I put the data into a database (Cockroachdb).