Maximum Availability Architecture – Oracle’s industry-leading set of database high availability capabilities

Out with the Old, In with the New: Database Platform Migration Done Right

Tim Chien
Senior Director of Product Management

Migrating Oracle databases from legacy infrastructure to newer platform and hardware is a must-have skill in the DBA's toolkit. The benefits of faster performance, higher resiliency, and cost savings with a newer platform - e.g. Oracle Exadata - must be weighed against the cost and effort of the migration, not to mention impact to database availability during this time. When large numbers and volume of databases are involved, the task becomes even more challenging, with application dependencies often spanning multiple databases which need to be moved as a single unit. So what is a viable approach?

A popular option for migrating large databases across platforms with minimal downtime is transportable tablespaces, where user data files are copied or restored from a full backup while undergoing endian conversion to the destination platform format. This is followed by periodic incremental backups on the source database which are applied to the destination database in succession to keep it in sync with the source, until finally, a Data Pump export is taken of just the metadata for the user tablespaces followed by import of the dump file into the destination database. In this way, the migrated data can be made available quicker than performing logical import/export of the entire data set. Furthermore, the source database stays read-write and available to users during the entire process, with exception of the final phase of the migration where a final incremental backup is taken on the source in read-only mode and applied on the destination to bring it up-to-date, followed by metadata export and import operation. Net-net, the total source downtime during migration can be reduced from several hours or more, to two hours or less.

Even better, when combining this migration approach with the Zero Data Loss Recovery Appliance, not only is downtime reduced, but there are additional benefits compared to using traditional storage:

  1. A new full backup is not needed prior to migration. With the Recovery Appliance, a full backup is only performed one-time, then incremental forever. The system produces space-efficient virtual full backups that can be restored as a normal full, including for migration purposes. These change-only backups results in significant time and storage savings, especially for large database sizes and/or number of databases.

  2. Virtual full restores eliminate the need to restore and apply incremental backups, as with traditional backup storage. The virtual full references all needed data file backup blocks and directly creates the needed physical full backups requested by RMAN restore. This reduces the overall migration window, and again, is especially key for large database environments.

  3. Every virtual full is regularly validated by the appliance to ensure success of future restore operations. Traditional storage cannot perform such checks on database backups.
  4. Built on the performant and resilient Exadata hardware platform, the appliance is designed to meet backup, restore, and migration demands for 100s-1000s databases, serving as a highly scalable, centralized 'data movement engine' across the enterprise.

More information on this innovative migration approach using the Recovery Appliance can be found in this presentation, MOS note, and demo of the migration steps. See Energy Transfer and METRO customer case studies who used this approach to move databases off legacy platforms to Exadata Cloud@Customer.

As always, let us know of any questions or feedback in the comment field below.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.