cancel
Showing results for 
Search instead for 
Did you mean: 

Deduplication and AVVI integration

massimiliano_2
Level 4
Partner Accredited

Hello there,

when deploying BE 2010 R3 for use with AVVI and deduplication, as per your experience does it make sense to create an initial "gold" full backup to a dedupe folder that will not expire/be overwritten then perform deduped full backups forever ?

Can anyone please share som field experience ?

Thanks and Regards,

Massimiliano

3 REPLIES 3

teiva-boy
Level 6

It doesn't matter.  BE will retain the blocks of data it needs, regardless of the append/overwrite periods.  So you do not have to pre-stage anything and keep it "indefinitely."

You also if capable (vSphere 4+ and HW level 7 VM's) would want to implement Full/Incr or Full/Diff AVVI backups and not just fulls everyday from a capacity standpoint.  This would would ensure that the index database doesn't grow too large, as well as not grow the dedupe folder larger/faster than anticipated.

InsentraCameron
Level 4
Partner Accredited Certified

Hi Teiva-boy,

From what you are saying then, the key thing disadvantage to doing continual fulls is the indexes will grow to quickly and be more of a hassel to manage. I have wondered in the past why people wouldn't do full backups forever.

teiva-boy
Level 6

The dedupe process works like this (mind you this is very high level):

  1. Scan the data on the volume
  2. Break the data into fixed segments
  3. Assign each segment a unique MD5 hash (UID)
  4. Update the database with these UID's
  5. Compare what the client sent to whats in the dedupe database via UID look-ups
  6. Tell client to only send the unique data segments from the volume.

 

So here is where things get interesting from what I just wrote...

  • You are limited by HDD read speeds in step 1.
  • #4,5 require fast disk, CPU and enough memory to read and update the dedupe database.
  • #3,4 will grow the dedupe database over time, prematurely.  If you continue to send FULL's, the database grows faster than normal, affecting the 2nd bullet point I mentioned above.

While you could do fulls all the time, it's not recommended, and you should just continue to do backups as you did in the past with a traditional full/incr or full/diff methodology.