Oracle 11g Total Recall/Flashback Data Archive Oracle 11g New Features Tips by Donald BurlesonJune 29, 2015

Oracle 11g New Features Tips

Tracking transactional changes on tables over its lifetime by using FLASHBACK DATA ARCHIVE

Also see Oracle Total Recall Tips.

More and more, legal regulations such as Sarbanes Oxley and Basel II are challenging all kinds of organizations by enforcing strict change control of customer data. These obligations demand that a history of all changes to customer data must be maintained.  Nowadays, companies are commonly required to retain their data for periods of 5 or more years, and must be able to review historical data almost in real time. 

One common approach to this problem is the implementation of essentially home grown data management systems which integrate the maintenance of the history and archiving in business logic. The application has to keep track of data changes which can make applications terribly complex, and create difficulties in applying upgrades to the application.

The approach of using PL/SQL triggers for tracking changes can have a massive impact on the performance of an application. This is because it is compiled source code, stored in the data dictionary, which needs to execute again and again. Another drawback of such a solution is that there is no central interface for the management of those triggers.

Most of the flashback functionalities rely on UNDO data which will error out with an ORA-1555 snapshot too old if the required before image cannot be read any more.  ORA-1555 occurs because it has already been overwritten in the undo tablespace. Furthermore, it is very unlikely that the old values can be reconstructed over a longer period of time, like months or even years, from data stored in the undo segments.

Luckily, this UNDO data time issue can now be remedied.  With the flashback data archive functionality of 11g the Oracle database is capable of automatically tracking transactional changes to data over very long periods. Oracle achieves this by storing UNDO information in special segments within dedicated tablespaces.

The usage of this feature is completely transparent for the application and the end user, who can view historical data from the flashback archive seamlessly with regular SQL statements.  This is done by utilizing traditional flashback functionalities, such as flashback query, flashback versions query, flashback transaction query, etc.

The historical information in the flashback data archive ages out automatically and Oracle automatically purges it after the specified retention period has exceeded. However, the flashback data archive provides the DBA with a central interface for the management of historical data and change tracking. 

The technology behind flashback data archive

With every Oracle 11g database startup, the new flashback data archiver background process, FBDA, is automatically started. This is what generates and archives the historical data.

Transactions encounter very little performance impact from flashback data archiving because Oracle only marks DML operations as candidates for archiving.  A special background process then generates and archives the history information asynchronously for tables enabled for flashback archival.

Oracle also automatically compresses the internally used history tables and partitions them based on a range partitioning scheme. The partitioning and compression of the history tables is fully transparent and does not require any additional administrative intervention.

A flashback data archive consists of at least one tablespace, and can span multiple tablespaces. It is possible to add a new tablespace to a flashback archive at any time.