MAT (Powered by Project Shift)

Those of you who joined us at any major event recently have hopefully come across a preview of some cloud-virtualization conversion technology that we’ve been working on here at NetApp. We first shipped the capability over two years ago, but it’s taken us this long to truly understand how customers see the problem and how best to meet the needs of the market.  What am I talking about?  Hypervisor mobility and NetApp Project Shift of course!  For those who haven’t been along for the ride here’s the abbreviated version.  Project Shift is a collaboration across all divisions here at NetApp to enable our customers to quickly shift a workload from any hypervisor to any other hypervisor (as well as physical to hypervisor or the other way around).  We do this by leveraging Data ONTAP and our deep understanding of virtualization. 


More specifically we use FlexClone to clone the data blobs from one format to another writing any unique differences as needed.  For instance, when shifting from VMware to Hyper-V, we clone from the flat VMDK into the new VHD/VHDX writing any headers and metadata as needed.  The end result is we’re able to take a VMDK on an NFS data store and shift it to a VHD/VHDX in seconds!   More importantly we’re doing so in an incredibly efficient manner as we’re only writing the difference in the file format not the contents of the virtual disk itself.


Hopefully you’ve had a chance to see the concept demo we released a little while ago.   I myself have been giving live demos of Shift since March. 

Now if you attended any of the major events we’ve previewed this at this year, then you probably already knew all of this.  However what you didn’t know was all this while we’ve been feverishly working with Microsoft to help bring this technology to meet a very specific need in the market today.  I first showed Shift to Microsoft two years ago and they thought it was really cool, however there wasn’t an immediate opportunity or use case.  That all changed this past April, when Microsoft released the Migration Automation Toolkit, or MAT.  I first saw MAT in action at MMS, and I was very impressed. Microsoft had built a tool to solve the largest problem facing their Hyper-V customer; getting their existing infrastructure off of VMware!  It did almost everything customers had told us they needed.  It handled the inventory and evaluation process, uninstalled VMware tools, and created the destination VM.  One problem though, the conversions themselves are very time consuming and require additional storage and network resources, in many cases a 100% increase in storage capacity .  This isn’t a technical problem it’s a case of simple physics.  The MAT must copy the source VMDK and stream the contents of the file to the destination host.  All in all I was impressed but wanted to know more so before leaving MMS I was able to meet MAT’s author Mark Gosson.


It was during this meeting that as Mark puts it we had our ”chocolate peanut butter cup” moment.   Here was Microsoft with a conversion toolkit, “rich and fantastic as peanut butter” that does end to end migrations today, and there we were with a decadent chocolate-like technology that can flip the VMDK in seconds. We recently got together to see how hard it would be to put these two great flavors together.  At which point Microsoft delivered a dream team of automation migration specialist in Mark Gosson and Michael Greene.  Between the three of us, we’ve integrated some of the conversion technology of Project Shift into the MAT.  Along the way we extended MAT to include the ability to migrate the network settings as well.  We’re calling the resulting technology MAT (Powered by Project Shift).


What does all this mean, well if you’re a NetApp customer and are interested in migrating a VM from VMware to Hyper-V, Microsoft now has a tool that can accommodate that migration in about 5 minutes!  If you’re not a NetApp customer and are in need of a shorter SLA to make a migration possible then we included the wizards to easily configure a new NetApp Controller to facilitate the migration itself.


If you’re not a Hyper-V customer, but are interested in the technology itself, perhaps to facilitate similar mobility in other directions, then we encourage you to keep NetApp on your short list as we plan on announcing the future direction of Project Shift. 


Until then Happy Migrating!





Are the vms then running off the flexclone on hyperv, or is the flexclone the back up before the changes to the file formats occur?

Also if your using nfs currently how do you get that onto a hyper v cluster? I'm assuming you need to be 8.2 c mode and share the nfs as smb 3.0 on the hyper v cluster?


Yes this solution requires cDOT because we need to share the same volume as both a NFS mount and SMB3.0 CA share.


Neither we're leveraging the SIS engine behind FlexClone to create a net new file that is backed by the physical blocks from the source.  Think of it as pre-dedupe, instead of copying the data and then letting SIS remove the duplicate blocks, we simply never write the redundant blocks in the first place.  In ONTAP this technology is commonly referred to as FlexClone, but we're not using it in the typical volume cloning workflow.


I have not seen in "Guest operating systems supported for conversion" Red Hat system.
Can we convert Linux VMs?


we're new in the shift-world!

Can you tell me the system requirement? OnTap version? - ClusterMode required?
And it is possible to shift from 6.x Xen HV to VMWare 5.5?

it would be possible for a beginner to get an example script for shifting?

As per comments above - yes, cDOT is required (sadly).

And from what I know, currently it's VMware to Hyper-V only, and only in one direction

ok, it work's from Xen to VMware - with the NetAPP PS convert.... - significantly faster than the converter from VMware


Hello people

i was thinking on using the tool to convert physical disks (LUNs) to VHDs (instead of disk2vhd) for Microsoft envs providing that my scenario  would work.

I know it does not.

But, would it imply too much effort to add this feature?

could it be added?

Thxs & Rgds

It is stunning to see the cloud technology. It is really getting better and better and introducing more and more options for the virtualization technologies. But I do not think there would be hundred percentage of transparency between Red Hat and Microsoft.

Skechers clogs


Article from Keith about NetApp Shift might be interessting as well:


That's awesome.

I had a look on the scripts that does such migration (MAT4SHIFT) and there is something weird:


This line 521

Get-Content ('{0}\{1}' -F $NFSPath, $HardDisk.replace('/','\')) | ?{$_ -match '\"(?<file>\S+.vmdk)\"'} | %{



My question is: Why the MAT scripts that are on a Windows machine needs to access to a NFSPath ? Should the content of the variable $NFSPath  be a real NFSPATH or the variable name is misleading and the get-content is performed on a CIFS share anyway ?


Thank you