When you run the %DI_JOB_RO JOB_EST task during the weekly batch processing, a significant amount of real time might be spent processing chain effects. The slow-processing problem occurs with SAS^® Regular Price Optimization or SAS^® Promotion Optimization. During chain effect processing for each model group, queries are run that identify other model groups related by hierarchy changes.

You can identify this issue by searching for the string "di_job_chain_effect_nodes finished successfully" in the <ddmonyyyy_hh_mm_ss>_di_job_est.log and note the total duration of all occurrences. The following example indicates that almost 7 hours were spent processing chain effects:

 
di_job_chain_effect_nodes finished successfully - started at 31MAR12:10:27:40, and finished at 31MAR12:10:29:00
...1350+ other occurrences...
di_job_chain_effect_nodes finished successfully - started at 31MAR12:17:01:21, and finished at 31MAR12:17:01:21

When all model groups are being calibrated, chain effect processing is not needed.

There is no workaround for this issue.

Click the Hot Fix tab in this note to access the hot fix for this issue.

Once the hot fix is installed, a new option for %DI_JOB_RO is available. The ENABLE_CHAIN_EFFECT option enables you to control whether chain effect processing takes place. With the default value, ENABLE_CHAIN_EFFECT=1, chain effect processing runs only if data_refresh=1 and model_group argument value is not -1. When ENABLE_CHAIN_EFFECT is set to 0, chain effect processing is disabled regardless of other arguments' values. When you are processing a subset of model groups, specify ENABLE_CHAIN_EFFECT=1.

Operating System and Release Information

A fix for this issue for SAS Merchandise Intelligence 4.2_M2 is available at: