The back-end data is a single Hive record set. That record set includes columns that were joined within Hive from Hive tables that I don't have access to. Anyway, that record set will probably change every 10 minutes or so (meaning, there will be new rows added every 10 minutes or so). There will eventually be millions of rows in that record set. I am wondering, what is the best approach to partitioning the tabular model, from an architectural perspective?
↧