Hi All,

A customer of ours was running a once a year EIM process to load accounts into s_org_ext. They complained that the processing time was 4x longer than normal. We looked and noticed for S_ORG_EXT the inserts into S_ORG_EXT were being done row by row (RBR) with bind variables instead of the normal set based processing where the insert is driven by a select from the EIM table. Our batch size is 10K so instead of one insert that inserts 10K records we we doing 10k inserts instead of just the one.

We know if transaction logging is enabled for Siebel remote or pre IP 17 local databases this is the expected behavior. The only other thing that can cause this behavior is if your specify and EIM IFB option to force transaction logging. Neither of these conditions were true so we we really baffled by what could be causing the behavior.

We gave up on understanding why we were getting the slow RBR inserts with bind variables and decided to just try to tune these individual inserts which were slow at 0.6 seconds each. To try to make the individual inserts run faster we decided to add ONLY BASE COLUMNS to our IFB file. The number of columns was just under 60 where without the only base columns there had to have been closer to 1,000 columns in the insert statement.

To our surprise once we added the ONLY BASE COLUMNS option to our IFB file the RBR behavior went away and was replaced by a select from the EIM table. So it looks like once the number of columns being loaded into a base table exceeds some limit EIM will revert to RBR processing instead of the normal set based select from the EIM table to insert all the row for the batch in one go.

With this improvement the job now runs way faster than it ever has!