There are 3 settings which control Archiving in PAM:
Minimum Days of Process History = Age in days before a completed instance will be archived. The timer starts upon completion of the process instance.
Option to purge Archive Data = Indicates to PAM if and when an archived process should be purged.
No of Days to Keep Archived Data = Indicates how long an instance will sit in the archive before being purged.
This is documented in Tech Doc TEC596125
This is complicated somewhat depending on your integration as PAM makes an API allowance for you to set the AutoArchive flag when launching an instance.
For example, all instances launched from Service Catalog are set with AutoArchive = False. This means that no instances started from Catalog will ever be moved through the Automated Archive process of the Orchestrator and can only be moved into the archive by a manual act of selecting the instance, right clicking, and choosing Archive from the Operations tab.
Beginning with the 4.2 SP02 release our development has performed a significant reworking of the archiving sub-system within PAM. In addition to making the archiving sub-system more robust, one aspect of this work is to make sure the runtime tables stay lean by immediately after a process completes and the roll-up is performed (consolidating the multiple rows that represent a running process into a single row), the resultant row is moved into the archival table.
Prior to 4.2 SP02 an instance would complete, be flagged for archiving, and would sit until the designated ‘Minimum Number of Days of Process History’ had expired, then the multiple rows that make up an instance run would be rolled up into a single row, then moved into the archive tables. In 4.2 SP2 this is done as soon as an instance run has completed.
This distributes this work across all hours, and assures that even if a Process has been excluded from the Auto-Archive policy that the Process row will not stay in the run-time table.
There will be no visible change for the PAM end users. The Completed process will still logically be un-archived and will be returned in ‘current’ searches on the Operations tab, even though the physical row will reside in the Archive table. When archival processing is run, those Processes that have Auto Archive enabled, and have met the other criteria will then be flagged in place as having been Archived.
Purging will work as before, and Archived rows that match the Purge policy in effect would be removed. The rows not marked as Archived will of course never be purged.
The bottom line is that these changes will make it so that even if Processes that are started with the flag that excludes them from being Auto Archived are never archived, this will not impact PAM performance is any significant way, as even a huge number of rows in the Archive table would only have minimal performance effect.