Deployment Planning Guide > Data Integrity and Capacity Planning >

Database Recovery Planning

Follow the RDBMS vendor's recommendations on configuring the database for recovery in case of data corruption, hardware failure, or disaster.

IBM DB2 Recovery Planning

The transaction log should be mirrored to guarantee database recovery in the event of a single device failure. The instance home directory must be mirrored, if resources are available. Hardware or operating system mirroring generally provides the best performance.

Oracle Recovery Planning

Many companies today use RAID storage systems that make Oracle online redo log mirroring unnecessary.

If your organization does not use RAID storage systems, you should, at a minimum, mirror the redo log, as this is essential when a database goes through crash recovery.

Also, when redo logs are mirrored at the RAID storage system level (usually RAID1 or RAID0+1), there is usually no need to mirror them at the Oracle level, since the RAID controller assures that these volumes can always be recovered. Mirroring at the RAID level usually improves database performance (especially beneficial for read operation).

If you have the resources, the Oracle control files should be mirrored as well. Otherwise, you can put the Oracle control files into a RAID 5 device as it is not heavily accessed and disk performance is not a concern. The information it records, though, is very critical for the Oracle database. Any updates to the control file—for example, the current System Change Number (SCN) or transaction tables—ripple across all members of the control file specification.

Deployment Planning Guide