How can I set up Hyper V to get the best saving with Dedup?

[#1]

I’ve been working on using Dedupe with Hyper V virtual servers, but I’m only seeing about 10% space savings. The documentation I’m reading says typical deduplication storage savings for Hyper-V environments is 60%. I looked for a document that detailed the best practices for using Dedupe with Hyper-V but all I could find was configuration documents. I listed out some of the questions I have. Can you please take a look and let me know what will give us the best performance and the most savings with Dedupe. Also, I read about a tool called “Space savings estimation tool” that will allow us to see how much savings we can expect to see with Dedupe. Is there any way we could get that tool. It would be nice run that on our volumes before enabling Dedupe so we can evaluate what volumes would have the most benefit.

·         Are we better off to run multiple virtual servers on 1 volume, designate a separate volume for each virtual?

·         If we have multiple drives, will we see a significant performance improvement by directly connecting the virtual server to the Netapp rather than using a VHD file?

·         Is Dedupe able to gain space saving from within a Windows Lun?

[calvarez]

Perhaps you can find useful details in the NetApp Hyper-V Deployment Guide at http://media.netapp.com/documents/tr-3733.pdfhttp://media.netapp.com/documents/tr-3702.pdf.

I will leave the Hyper-V-specific questions around best practices to the Hyper-V experts, but I can lend some insight into the dedupe part of your questions.

First, with regard to the Space Savings Estimation Tool (SSET), you will need to contact your local NetApp account team, or certified NetApp partner.

Next, the question about multiple servers within a single volume or each having their own volume.  From a strictly dedupe perspective, dedupe works at the volume level.  So if multiple virtual servers are stored within a single volume, then there will be more duplicates within that volume, and deduplication will be even more effective.  The Hyper-V experts will need to tell you how to do the proper sizing to determine how many servers should be within the single volume.

And in regard to whether dedupe will provide savings within a Windows LUN, dedupe works at the block level and will work in both SAN (LUN) and NAS environments.

[erick_moore]

I haven't used Hyper-V, but based on what I know about it and the NetApp you are going to want to build your LUNS for Hyper-V in the same volume.  Just think of a volume as a logical container.  LUNS are nothing more than files to OnTap so you will have no problem storing multiple LUNS in the same volume.  As was pointed out at this point De-Dup is only supported at the volume level.  Because of this data you want to de-dup needs to be in the same volume for maximum savings. Here is a real high level setup. 

1.  Build a 100GB volume

2.  Build 4, 20GB LUNS in that volume (make sure to thin provision the LUNS)

3.  Install your OS's, configure, etc

4.  Run asis on the volume created in step 1.

Windows will not be able to see the space you have reclaimed using De-Dup, but you will have more room in the volume for things like additional LUNS and snapshots.  If you will be adding more LUNS in the volume (maybe you have 6, 20GB LUNS on that 100GB volume after thin provisioning and De-Dup "over provisioning the volume"), make sure you set the volume to auto-grow, and/or snap auto-delete.  If a volume containing a thin provisioned LUN runs out of space it will offline all LUNS in that volume.  The document link provided does a good job of explaining all of these functions.  I would also recommend the Data OnTap Fundementals class if you are new to the NetApp.

Thanks,

E-

[peluso]

Hi -- I've also passed this along to our Hyper-V subject matter experts for more info....stay tuned!

Terri

[lfreeman]

FYI here is a paper that goes into more detail about setting up NetApp storage up with Hyper-V

http://media.netapp.com/documents/tr-3733.pdf

Larry

[amiller_1]

In short.....

• put LUNs with similar data inside the same volume (since dedup is at the volume level)
• use thin provisioning to use the space freed by dedup (this is essentially the approach you use for any LUN/dedup data)