RMAN Proxy Copy for better dedupe?

[D_Thomas]

Our RMAN dedupe rates are pretty pitiful. We are seeing rates of 57% on full backups of databases that only append small amounts of data relative to their size - the DB is 1TB, but we are only adding 5 GB to the size each week.

I have seen that the proxy copy method will help with this, but it seems sketchy to me to say the least. I know very little about Oracle, but I need to help work with the DBAs on changing this if it will indeed fix this dedupe issue.

Needless to say, this is only one problem we seem to be having with RMAN, it has been very difficult to implement on any sort of consistent basis - is this true across the board?

Thanks,

DT

[Mark_Solutions]

Databases are often difficult when it comes to de-dupe

I often make sure that compression is set all round for logs/ incremental backups which does seem to help

There is however an Oracle De-Dupe Best Practices guide, not sure if you have seen it?:

http://www.symantec.com/docs/DOC3534

[D_Thomas]

I understand that they are hard to dedupe, but this seems exceptionally poor for a database that barely changes. I have seen the best practices doc and that is what we are using.

I had read that using the proxy copy provides a more consistent stream of the data, which makes sense. It was a comment from a Symantec employee in another thread. However, It looks like a more involved setup. I was hoping someone might have some guidance around setting this up vs the traditional rman backup. The oracle doc mentions something about downloading a different binary from netbackup, which already puts me a little off.

[Marianne]

The oracle doc mentions something about downloading a different binary....

Which version of NBU are you using? The EEB's are only applicable to older versions of NBU (7.0.1 and earlier). No EEB's needed for 7.1.x and 7.5.x.

Have you tried normal stream-based backups with FILESPERSET=1?

Extract from Oracle De-Dupe Best Practice guide:

For stream-based backups, Symantec recommends that you specify FILESPERSET=1 for all Oracle database backups.When FILESPERSET=1 is specified, Oracle generates the backup set identically each time, with the same data from the same files in the same sequence each time the database is backed up. This uniformity ensures better deduplication. In addition, when FILESPERSET=1 is in effect, Oracle does not perform multiplexing, so Oracle includes only one file in each backup set. If FILESPERSET is specified with a number other than 1, Oracle groups files together unpredictably and deduplication rates suffer.

Increase number of channels when FILESPERSET is changed to 1.

Also take note of this Note in the Best Practice Guide:

Note: Make sure that Oracle optimization, encryption, and compression are

disabled for the entirety of the database backup. Make sure to check specifications

outside of the RMAN backup script, too.

[D_Thomas]

Yes, we are using filesperset=1, and following all the other guidelines. We mostly get decent deduplication, but on some file parts, typically the largest ones, we get poor deduplication - as noted above - 57% typically.

There are some comments here and there that proxy copy provides a better dedupe rate, so I was curious if anyone has had any experience with it. I guess we just have to live with it.