Siebel Database Upgrade Guide for DB2 for z/OS > Basic Database Preparations for a Siebel Upgrade >

Reviewing EIM Table Partitioning and Data Distribution

Upgrades: All upgrades.

Environments: All environments.

Partitioning EIM table spaces can improve EIM processing performance. However, make sure that you check that the partitioning keys you are using distribute the data evenly in the table space partitions.

Siebel Business Applications provide a partitioning column, PARTITION_COLUMN, designed specifically to produce even data distribution. This column is populated with data using a BEFORE INSERT trigger to generate the partitioning value for each row.

In releases of Siebel Business Applications before 8.0.x, the trigger that is used to populate PARTITION_COLUMN in the Siebel prepartitioned tables is based on the ROW_ID column, or another column, from a parent table. In Siebel releases since 8.0.x, the BEFORE INSERT trigger generates a random number between 00 and 10 and uses this number to populate PARTITION_COLUMN.

When you upgrade from a pre-8.x Siebel CRM release to Siebel CRM 8.x, the PARTITION_COLUMN trigger is not changed. For Siebel prepartitioned tables that are partitioned based upon PARTITION_COLUMN using the 7.x release trigger, you might have to modify the partitioning values you use to achieve balanced partitions.

The mechanism by which EIM generates the ROW_ID can result in an uneven distribution of data if you use EIM with one of the default partitioning schemes to import data into a base table. When EIM imports data into a base table, EIM automatically generates the base table ROW_ID by concatenating the prefix with a unique number; for example, 1-SB3-123. EIM relies on database functions to make sure that each number is unique under the same prefix.

Uneven data distribution occurs because one of the default partitioning approaches uses the last two characters of ROW_ID as the partition key for partitioning of base tables; these characters can be letters, numbers, or a combination of letters and numbers, but the last two characters in a ROW_ID generated by EIM are numbers.

If your storage control file is to use partitioning keys that reflect the nature of your data, consider your data distribution with regard to your EIM process before you begin your database upgrade. Then, modify your partitioning keys accordingly. Alternatively, you can repartition the table after an uneven distribution has occurred.