Oracle Deduplication and Compression Setting

Article ID: ORA0001 Oracle Deduplication and Compression Setting


Oracle Deduplication and Compression Setting


With Oracle Database Agent you should not use deduplication with compression enabled. This will result in a lower deduplication rate. This is caused by how RMAN sends the backup data.

For Oracle (and Oracle only) we recommend to disable the compression when deduplicating the backup data. We understand this will result in a bigger baseline on the initial full backup. But the savings with subsequent backups will be much higher. 

However, you need to consider retention before turning compression on or off. If you have short retention (e.g. 4 weeks) turning compression on may give you less deduplication, but the space savings in the compressed initial baseline may more than offset the additional data written by subsequent Fulls. On the other hand for longer retention (8+ weeks) turning compression off may be a better option since the extra space consumed by the initial baseline may be offset by the space savings across Fulls.

Another factor to consider is DASH copy. You should always have a secondary backup copy.  With deduplication you have the option of using DASH copy. If bandwidth is a consideration for creating the secondary copy you may want to disable compression since that will result in less data transferred over network during DASH copy.  Again, this is a tradeoff of network performance versus consumed disk space.

There are other ways to improve deduplication reduction for Oracle, even with compression on:

  • Use incremental storage policy without dedupe and point incremental/archive log backups to the incremental storage policy. Since most incremental jobs contain unique data that does not dedupe, there is no point in wasting cycles trying to do dedupe.
  • Reduce the number of streams for Oracle. The higher the number of streams, the lower the reduction across Fulls. Oracle admins are reluctant to decrease the number of streams as this tends to reduce performance of RMAN backup. However, it is worth pointing out that source side dedupe reduces around 70-90% of data transfer over the network, hence improving backup speeds by 50% in many cases. As a result, source side dedupe backups with 2 streams can potentially perform as well as 4 streams without source side dedupe. This assumes, of course, that you are performing LAN based backups.