Skip to Main Content
Return to Navigation

Understanding Archiving Techniques

This section discusses:

Business Requirements Analysis

It is important to devise a business strategy before archiving the data. First, you must identify the tables that you want to archive. This includes identifying all of the parent and child tables associated with the tables. Failing to identify all of the related tables can cause corruption to the database. Next, you must know exactly which data to archive. It is important to recognize which rows are safe to remove from the online tables. Remember to remove only the data that is not required to maintain the day-to-day business and reporting.

Consider PeopleSoft General Ledger as an example. General Ledger contains the greatest amount of data to be archived because it is the module where the majority of reporting is required. There are two sets of data types that need to be maintained: balance information and transactional information. Balance information is retained in the ledger records. You might require balance information for online and reporting purposes to be available for a three-year period. On the other hand, transactional data is maintained in the journal header and line tables. Suppose that you require only one year of transactional data to be retained in the system for online purposes, but three years to be retained for reporting purposes.

Any data beyond the above time frames for balances and transactions can be archived and is only be accessed through reports. The data can be archived to history tables. If data were to be archived into history tables, the data would still be available online for reporting purposes. However, you could not view it through standard PeopleSoft Internet Architecture pages without special configuration. In addition, reports would need to be modified to access the data in history tables. Moving archived data to secondary storage devices is generally used for long-term data retention. This option is preferred for data that is rarely retrieved, and secondary storage devices are usually used to satisfy legal requirements.

Commits

By default, the Archive Selection, Remove from History, Rollback, and Delete processes issue commits after each record has been processed unless Row-based processing or Unit-of-Work processing have been specified.

Enhancing Performance

For better performance and increased speed during archiving processes, consider dropping the indexes before inserting data from online tables into history tables.

Index Considerations

Index considerations include:

  • Database differences.

  • Non-unique indexes.

Platform Considerations

The database platform may have a limitation on the number of columns that an index can contain. Some have a restriction of 16 columns for an index. If the table that you want to archive already has 16 keys, then you can't add other keys (PSARCH_ID and PSARCH_BATCHNUM from PSARCHIVE_SBR sub-record) to the corresponding history table.

To solve this problem, you can create the history table with the PSARCH_ID and PSARCH_BATCHNUM as non-key fields.

Non-Unique Indexes

The SQL generated by the Data Archive Manager assumes that index keys identify unique rows. Therefore, the base table of the base object must have unique indexes.

Data Limitations

For Oracle databases only, due to platform and meta-SQL restrictions, Data Archive Manager does not support archiving of records with LONG, IMAGE, or ATTACHMENT columns if you have not performed a data type switch. If you have performed a data type switch, there are no limitations. The selection process (inserting data from the online records to the history records) will result in the loss of the long, image, or attachment columns in the history record.

However, this restriction applies only to templates archived using set-based processing. Long, image, and attachment data are archived to history records (and back to the transactional records) if the template is archived using row-based processing.

Note: This potential limitation applies only to Oracle databases. No other databases are affected.