Forum Discussion

kwakou's avatar
kwakou
Level 4
9 years ago

Client side dedup

Dear Experts, i need your help to understand the following.

I have created an oracle intelligent policy and it's running fine.

I measured the elapsed time: it took me 58' to backup 500 GB of DB.

I then decided to enable client side deduplication and choose the option " always use client side deduplication"

the first time i ran it, i got the backup taken in 1h02 '

And Surprisingly the second time when i was expecting impprovements, it took almost the same time !

But from the activity monitor , i can see that the client direct dedup is working. and even that the dedup rate is 100% (this is a test environement and no change is being made between my changes, also no other activity is running a part from my test)

Other fact : the number of files backed up are the very same from each attempt to another.

PS: kindly note that my oracle policy is using 4 streams.

How can we explain the fact that we r not getting any reduction in the backup time ?

 

 

  • Please could you show us the numbers that you interpret as evidence of this behaviour?

    Maybe... the very first RMAN test backup did also actually dedupe extremely well against existing previous (test and/or prod) RMAN backups, and so, your 1st, 2nd and 3rd test backups would indeed then exhibit very similar behaviour.

    Remember, the KB count in activity monitor is not reduced by 'client side' dedupe.  The activity monitor KB count remains the same, as this is the "effective size" of the backup job, if it were to be re-hydrated, i.e. it is the size of data before de-dupe occured, it is not necessarily (although it can be) the size of data sent over the wire.

3 Replies

  • As far as I know the data is received from rman, so rman has to provide Netbackup with the entire stream and the client will then deduplicate before sending to the media/storage server.

  • Please could you show us the numbers that you interpret as evidence of this behaviour?

    Maybe... the very first RMAN test backup did also actually dedupe extremely well against existing previous (test and/or prod) RMAN backups, and so, your 1st, 2nd and 3rd test backups would indeed then exhibit very similar behaviour.

    Remember, the KB count in activity monitor is not reduced by 'client side' dedupe.  The activity monitor KB count remains the same, as this is the "effective size" of the backup job, if it were to be re-hydrated, i.e. it is the size of data before de-dupe occured, it is not necessarily (although it can be) the size of data sent over the wire.

  • Dear Riaan/Sdo

    sorry for the delay in my feedback.

    I was off the last 3 weeks. Let me go back into it and revert asap.

    Thanks