I have a DB-savvy customer who is concerned their silver/gold layer is becoming too expensive. These layers are heavily denormalized, focused on logical business entities (customers, claims, services, etc), and maintained by MERGEs. They cannot predict which rows / columns are going to be updated so ZORDER and partitioning have proved ineffective at limiting files to change and thus compute cost. They want to understand EDW patterns used by other large, sophisticated data engineering teams.
