dedupe on vm vs cifs

[bellows]

We have a CIFS share on our FAS2240 that is about 2TB in actual size.  We get an incredible dedupe rate of 66%, so used space is only around 900GB.  I would like to host this data from a Windows Server 2012 R2 vm instead of directly off of the Netapp CIFS.  My question is how much dedupe percentage will I lose by doing this?  I assume there is some overhead from putting those files inside NTFS on top of VMFS that would make dedupe less efficient.

[JGPSHNTAP]

Just curious, why would you do that? 

Why add another layer on-top?  Unless you say you need SMB 3.0 and cannot move to cDOT, I would leave it on netapp

[bellows]

Two reasons:

1. SMB 3 (as you said)

2. Veeam - I'm still having to use Arcserve for the CIFS shares.  I'd really like everything to be in Veeam as the single solution.

We are a single site, so snap-mirroring off site is not an option. We currently use Veeam backup to a Synology NAS that is replicated to a partner off-site.

[JGPSHNTAP]

So, you are going to use iscsi lun? or go down NFS route?

I've never actually thought about this question, now you got me thinking..

[bellows]

We use 8g fibre channel luns.  I've already created the VMDK in an existing data store.  Just haven't pulled the trigger yet until I figured out how much dedupe would be affected.

[JGPSHNTAP]

Well, Im curious as well..

Why don't you just robocopy the data over and don't swing the name and then just run a dedup job to see.. I have my suspicions but i will with-hold them for now