Several ways to adjust the conceptual schema:
- re-normalization: Consider alternative BCNF decompositions that better fit the queries in the workload.
- denormalization: Consider merging relational schemata to intentionally increase redundancy.
horizontal partitioning (separate rows)
- Each partition has all the original columns and a subset of the original rows. Tuples are assigned to a partition based upon a (usually natural) criteria. Often used to separate operational from archival data.
vertical partitioning (separate columns)
- Each partition has a subset of the original columns and all the original rows. Typically used to separate frequently-used columns from each other (concurrency hot-spots) or from infrequently-used columns.