Defining the data to load into an accelerator

Defining the data to load into an accelerator basically means selecting the proper tables. However, you also need to complete a number of related sub-tasks. This section lists all required tasks in the correct order.

Selecting the proper tables

In choosing the proper tables, you determine the data to be queried. For each Db2 for z/OS table that you select, a corresponding table is created on one or more accelerators. These tables are called accelerator-shadow tables.

Accelerator-shadow tables are not necessarily copies of their originals. Often, they contain fewer columns than the tables they stem from because columns with unsupported data types are excluded. Accelerator-shadow tables must be filled with data before they can be queried. This is done in a later step (see loading accelerator-shadow tables).

Important: The only type of Db2 for z/OS tables you can select are base tables. The following types of tables are not supported, and therefore cannot be selected:
  • Db2 for z/OS views
  • Auxiliary tables
  • XML tables
  • Materialized query tables (MQTs)

Checking the table size

When you select a table for an accelerator in the Add Tables wizard, your administration client displays the table size. This table size is calculated on the basis of the values in the CARDF and AVGROWLEN columns of the SYSIBM.SYSTABLES table. Thus the metrics come from Db2 for z/OS. Comparing this size with the available disk space on your accelerator, you can roughly assess how much space the table will occupy on your accelerator. However, Db2 for z/OS and the accelerator use completely different compression algorithms, so the size of a table in Db2 might differ from the size of the same table on the accelerator.