Batch-Processing controls

These controls are inherent in any batch-processing environment. Most of the organizations having stable system environment should have these controls.

Roll Backs and Commits

A transaction should commit itself (write onto the database), once all the processing to bring the database into a state of consistency is completed. If there is a failure anywhere, it should rollback itself. For example – Writing down the complete voucher details (including all Debits and Credits) before committing.

Today’s databases have sophisticated methods to drive these commits, and rollback options. Apart from consistency, the commit and rollbacks are also used for managing large processing jobs, for better performance management.

For large batches, the commit and roll back spans become bigger. This reduces the commit overhead, so to avoid doing commit after every transaction.

Batch Cut-offs

These controls assure that the batches do not start OR end before OR after the prescribed time limit. There are many reasons.

For example the Credit Card billing cycle batch to start by 9.00 P.M. (after most of the transactions are done), post which the card statement generation process to begin by 2.00 A.M. so that the spool files can be sent to the card statement printing batch by 5.00 so that the collation and stamping etc. is done by 8.00 A.M. to be picked up by the postage and shipping company.

Pre and Post batch-processing - Sequential controls

Every batch process has to have defined 'start criteria' and 'end criteria'. You would not like to run you month-end valuation batch, till the month end business transaction accounting batch has run.

Batch Failure Response Rules

These rules define the response of a system, as a failure or a defect is detected during the batch run. The rules could include:

• Skipping the record at which the failure happens and place it in the error log.
• Stopping at the record, where the failure happens, roll-back to the previous committ point, and wait for the problem to be resolved.
• Place a default value in the record, where the failure happened, so that processing can proceed. These default values to be corrected in the next processing cycle.
• Cancel the batch and roll-back to the start point of the batch and wait for the problem to be corrected.

In all the above we assume that a log is maintained around the errors encountered and action taken.