Deduplication with VMs

[Doug_Morrison]

Hi All,

Trying to find a workable solution for deduplication on our VM infrastructure.  Our environment is posted below:

Backup Exec 2014 SP2 installed on 2k8r2.

VMs are spread across 3 ESXi 5.1.0 hosts.

I initially created a dedup job that connected directly to the hosts to backup the VMs.  The throughput for the Full wasn't awful, averaging 1300MB, but the Incremental was terrible averaging 360MB.  I read here that backing up with this model is ineffective as changes to VMDKs are not seen granularly and so BE backs up the whole VMDK again:

http://www.robsteele.co/best-practices-for-backup-exec-deduplication/

So, I installed the agent on all the VMs and created a new job with 25 VMs added as physical machines.  Adopting this approach means I cannot use client-side dedup and the average job rate was 220MBs.

Just out of interest I created a job for a single VM and the throughput was great at 1800MB+ before I cancelled it.  The problem with this is that I won't be getting the full benefit of dedup as it will only dedup across individual VMs rather than the cluster.

My question is, what is the best practice for backing up VMs to dedup storage? 

Any help will be greatly appreciated and rewarded with Interent kudos points...

[lmosla]

Hi,

see if this helps https://www-secure.symantec.com/connect/sites/default/files/Backup%20Exec%202014_Deduplication%20Opt...

[pkh]

The site that you are referring to is hopelessly out of date.  Stream handlers for VM's are available for some time now and certainly for BE 2014, so you would not be suffering from the poor dedup ratio mentioned.

To backup VMware VM's, you cannot do client-side dedup because there is no client on the host.

Have you tried doing a backup to disk storage and see what the performance is like?

[Doug_Morrison]

Hi PKH, thank you for the reply.  I have tried a regular B2D job, which averaged 2500MB/Min for both the full and incremental jobs.  The full job is 1.6TB while the incremental is about 700-900GB.

The dedup storage is a LUN on the same storage array as the B2D storage area.

Any thoughts?

[pkh]

1. Make sure your media server has sufficient resources to handle dedup. See this document http://www.symantec.com/docs/HOWTO74446 2. Check that your disk array can handle the extra I/O's generated by the dedup process. 3. Check the job log and see whether it is the backup stage or the verify stage which is slow

[Gurvinder]

What if you only do full backups. How much time does a Full and Incremental take individually ?

If you do full backups , it is adding only the changes to the dedupe storage as we do check what fingerprints are allready present and only backup unique new segments, hence in a way only changes get added to your dedupe store.