Setting a checkpoint frequency

IMS records information for restarting interrupted operations by using checkpoints. When an operation is interrupted, reprocessing occurs from the point of system interruption to a forward continuation point. Using the status information captured during checkpoint, IMS can restore the content of the message queues and database changes.

Use one of the following methods to control the frequency of IMS internal checkpoints:

Specify the CPLOG= parameter in the DFSPBxxx member of the IMS PROCLIB data set.
Specify the CPLOG= parameter in JCL.

The value you choose for this parameter represents the number of system log records that are created before a new system checkpoint is invoked. When the increment exceeds the CPLOG= value, checkpoint processing begins. The default is 500,000. You can also use the /CHECKPOINT command to initiate a checkpoint. Application programs that are authorized to enter the /CHECKPOINT command can also initiate a checkpoint. The /CHECKPOINT command, used with certain parameters, shuts down your IMS system. Make sure that you understand the consequences of shutting down the system before you issue this command. After the checkpoint, use the /NRESTART command to restart the system.

Take frequent checkpoints to minimize the amount of time it takes to do a restart. The trade-off for taking frequent checkpoints (and efficient restart) is an increase in processing overhead.

You can change the CPLOG value by using the /CHANGE CPLOG command. To calculate a revised CPLOG value, use the transaction profiles and estimate, for a small peak processing interval, the number of system log events. The number of log records exceeds the sum of the following values:

(Number of input transactions) x 2
(Number of output messages) x 3
(Number of secondary transactions) x 2
(Number of database update calls for REPL,DLET)
(Number of database inserts) x 2
(Number of related index pointer maintenance calls) x 2
(Number of programs scheduled and terminated)
(Number of checkpoint calls from programs) x 3

To ensure an accurate estimate, your checkpoint frequency should be large enough to extend beyond the estimated response time for at least one of the longest-running transactions.

You can adjust the checkpoint frequency using /CHANGE CPLOG after observing the stability of the online IMS system. Use IMS Monitor reports to assess the processing overhead. The Region Summary report shows the total elapsed time and average elapsed time taken for checkpoints in the trace interval.