Prerequisities for adopting column-organized tables in partitioned databases

If you would like to introduce column-organized tables into a partitioned database, the database must meet the prerequisites for column-organized tables in non-partitioned databases.

If you are creating a new massively parallel processing (MPP) environment and plan to populate it with column-organized tables, follow the general system sizing and configuration recommendations for column-organized tables. Set DB2_WORKLOAD to ANALYTICS before you create the database. This automatically configures the database for columnar workloads.

If you want to introduce column organized tables into an existing MPP environment, follow the recommendations mentioned above. However, run the AUTOCONFIGURE utility with the APPLY NONE option to view the recommended configuration changes when running column-organized tables.

There are three approaches you can take to introducing column-organized tables into your partitioned database environments. These are the replacement, hybrid, or independent methods:
  • Replacement
    • Replace all traditional row-organized tables with column-organized tables.
  • Hybrid
    • Convert some of the row-organized tables to column-organized tables, or add new column-organized tables to an existing row table schema.
  • Independent
    • Introduce a new, independent columnar workload with no interaction with any existing workload or row table schema.